Semiconductor layer stack and method for producing same

Abstract

A semiconductor layer stack, a component made therefrom, a component module, and a production method is provided. The semiconductor layer stack has at least two layers (A, B), which, as individual layers, each have an energy position of the Fermi level in the semiconductor band gap, $E_F-E_V<\frac{E_G}{2}$
applying to the layer (A) and $E_L-E_F<\frac{E_G}{2}$
applying to the layer (B), with E.sub.F the energy position of the Fermi level, E.sub.V the energy position of the valence band, E.sub.L the energy position of a conduction band and E.sub.L−E.sub.V the energy difference of the semiconductor band gap E.sub.G, the thickness of the layers (A, B) being selected in such a way that a continuous space charge zone region over the layers (A, B) results.

Claims

1. A semiconductor layer stack comprising: at least two layers, a first layer and a second layer of the at least two layers each having an energy position of a Fermi level in a semiconductor band gap of $E_F-E_V<\frac{E_G}{2}$ applying to the first layer and $E_L-E_F<\frac{E_G}{2}$ applying to the second layer, with E.sub.F being an energy position of the Fermi level, E.sub.V the energy position of a valence band, E.sub.L the energy position of a conduction band and E.sub.L−E.sub.V the energy difference of the semiconductor band gap, wherein a thickness of the layers is selected such that a continuous space charge zone region over the first and second layers results.

2. The semiconductor layer stack according to claim 1, further comprising an extrinsic or intrinsic doping resulting in at least two impurity levels, each having an energy position, $E_F-E_V<\frac{E_G}{2}$ applied to the first layer and $E_L-E_F<\frac{E_G}{2}$ applied to the second layer, in a concentration such that on such infinitely thick individual layers, the Fermi level occupies the same energy position as the impurity levels with a tolerance of ±50 meV, wherein the thicknesses of the layers in the semiconductor layer stack are selected such that a continuous space charge zone region forms over the overall layer stack.

3. The semiconductor layer stack according to claim 1, further comprising an energy position of the deep impurity levels produced by the doping, $E_F-E_V<\frac{E_G.Math.0.8}{2}$ applied to the first layer and $E_L-E_F<\frac{E_G.Math.0.8}{2}$ applied to the second layer.

4. The semiconductor layer stack according to claim 1, further comprising an average Fermi energy position E.sub.F in an energy range $\frac{E_G}{2}-E_G.Math.0.2<E_F<\frac{E_G}{2}+E_G.Math.0.2$ in the space charge zone region.

5. The semiconductor layer stack according to claim 1, further comprising a doping with acceptor-like or donor-like dopants.

6. The semiconductor layer stack according to claim 1, further comprising one of the following dopants in the first layer alternating with a respective second dopant in the second layer in group III nitride semiconductors: iron and carbon; or carbon and a donor; or iron and magnesium; or iron and zinc.

7. The semiconductor layer stack according to claim 1, further comprising a sequence of at least two layer packets, including at least two layers with a Fermi level position, $E_F-E_V<\frac{E_G}{2}$ applying to the individual layer in the first layer and $E_L-E_F<\frac{E_G}{2}$ applying to the individual layer in the second layer.

8. A component module, comprising at least one component containing a semiconductor layer stack according to claim 1.

9. A method for producing a semiconductor layer stack comprising: providing a substrate in a system for the deposition of semiconductors; and applying a sequence of at least two layers, a first layer and a second layer of the at least two layers having an energy position of a Fermi level in a semiconductor band gap of $E_F-E_V<\frac{E_G}{2}$ applied to the first layer and $E_L-E_F<\frac{E_G}{2}$ applied to the second layer; selecting a thickness of the first and second layers such that a continuous space charge zone region of $W_A+W_B=\sqrt{\frac{2.Math.{\mathrm{.Math.}}_s}{q}\left(\frac{N_A+N_B}{N_A.Math.N_B}\right){\Psi }_{\mathrm{bi}}}$ over the first and second layers results, and D.sub.A≤W.sub.A and D.sub.B≤W.sub.B applies with W.sub.A and W.sub.B of the space charge zone and N.sub.A and N.sub.B being the dopant concentrations in the first and second layers A and B, ε.sub.S the dielectric constant, q the elementary charge and Ψ.sub.bi the built-in potential difference, which is identical to the energy difference of the Fermi level.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) The present invention will become more fully understood from the detailed description given hereinbelow and the accompanying drawings which are given by way of illustration only, and thus, are not limitive of the present invention, and wherein:

(2) FIGS. 1a to 1d show schematic views of an energy-location-diagram;

(3) FIGS. 2a, 2b, 3, 4, and 5 show energy positions of the bands relative to the Fermi level; and

(4) FIGS. 6a and 6b show the comparison of current-voltage characteristic curves.

DETAILED DESCRIPTION

(5) Because of the low charge carrier concentration present in the compensation of the excess charge carriers, the space charge zone generally extends over a range of several micrometers and thus results in a virtually shallow band profile with a slight modulation of the energy of the respective bands in each of the areas as shown in FIG. 1d or even less if the differently doped layers are chosen thin enough.

(6) Ideal layer thicknesses are dependent on the intrinsic background doping, the type of impurity, i.e. whether acceptor or donor, and the energy position. In individual cases, the optimal doping must be simulated or found experimentally. In principle, in a semiconductor having n-type background doping, the charge carriers are already captured for the most part by an acceptor seated in the upper half of the energy gap. The charge carriers that then remain in the conduction band are captured by a second acceptor in the lower half of the band gap. The resulting space charge zone is then so wide that in almost the entire layer stack, the Fermi level is close to the lower acceptor and therefore this layer has to be rather thin so as not to lower the Fermi level too much and on average to achieve a position of the Fermi level close to the middle of the band gap, so as not to induce any noteworthy hole conduction. But even without choosing the optimum thickness, as shown in FIG. 6, improvement in the layer insulation can be achieved since the strong depletion of charge carriers is presumably achieved locally. In addition, even without applied voltage at the layer stack, the position of the Fermi level also depends on how the adjacent layers are doped.

(7) For example, if a low, i.e. in the range of about 10.sup.16-10.sup.17 cm.sup.−3, but selective n-type doping by a shallow donor is chosen and said doping is to be compensated with a deep acceptor in the lower half of the band gap, it is advantageous to dope with a high concentration of the deep acceptor, well above the electron concentration, but only in thin layers. This results in a semiconductor that is completely depleted of electrons and has only a low concentration of holes due to the thin layers with the deep acceptor, in contrast to a layer doped only with the deep acceptor. These thin layers that are highly doped with a deep acceptor must be placed so close that the space charge zone regions that are formed overlap. In this case it is also necessary to dope with the deep acceptor at the beginning and at the end of the intended insulating layer stack in order to obtain complete insulation and no conductive edge regions. Such layers are advantageous for the switching behavior of components, since the re-emission of charge carriers from deep impurities, when the applied voltage changes, is minimized due to the reduced number of deep impurities and, at the same time, low residual conductivity.

(8) The doping according to the invention in the semiconductor layer stack can be done as doping with acceptor-like or with donor-like dopants, but also as doping with acceptor-like and donor-like dopants. It is true that implementation is possible only with acceptor-like or only with donor-like dopants and in some cases also makes sense, but theoretically a combination of one donor and one acceptor is ideal, since these better compensate for electron and hole injection, as is known from the Fe and Ti compensator in InP which are energetically very close to the middle of the energy gap and therefore, no adjustment according to the invention of the Fermi level by the combination of two dopants is necessary. If acceptors and donors are used, it is not easy to dope them in alternating layers due to the normally present background doping of electrons or holes, since deep donors do not capture or compensate for electrons and deep acceptors do not capture or compensate for any holes. For example, in the case of an n-type background doping, this can be achieved by co-doping the donor with a shallow p-dopant in a concentration above the electron concentration.

(9) Alternatively, a donor in the upper half of the band gap and an acceptor in the lower half of the band gap can be used. In GaN, the latter enables, for example, C to be a deep acceptor in a first layer and a deep donor without p-co-doping in the upper half of the band gap in a second layer. If a space charge zone forms over the area, the slight p-type conduction induced by the acceptor is then compensated by the donor in the second layer, although both energy positions are far outside the middle area of the energy band gap. Such a structure with a deep acceptor in layer A and a deep donor in layer B can be ideally implemented. In such combinations, acceptors in the lower half of the band gap and donors in the upper half of the band gap are preferably used.

(10) The structure or the semiconductor layer stack of the invention comprises at least two layers, that is a sequence of at least two layer packets including at least two layers (A, B) with a Fermi level position,

(11) $E_F-E_V<\frac{E_G}{2}$
applying as an individual layer in layer (A), and

(12) $E_L-E_F<\frac{E_G}{2}$
applying as an individual layer in layer (B), wherein it is advantageous to alternate these multiple times, i.e. to grow ABABAB, ABABA or BABAB, and also possible are layer sequences such as ABCBA or ACBADB, etc. having C and D arbitrary intermediate layers, which, however, may not prevent the formation of a continuous space charge zone region (110) due to their thickness and doping.

(13) According to the invention, a component module is proposed, comprising at least one component which contains a semiconductor layer stack according to the invention.

(14) In principle, the structure according to the invention is also possible only with shallow dopants or combined with shallow and deep donors and/or acceptors. This means a deep acceptor and a shallow donor and vice versa or only a shallow donor and a shallow acceptor, as shown in FIG. 5. In this case, multiple layers are necessary for an insulating effect, since only one p-n junction, i.e. one diode structure, is created with only one pair of layers A and B. At a lower Fermi level according to the invention, a p-n structure is also generally created, but the layers are then already so highly resistive that the current flow through them is greatly reduced and the diode characteristic would be only very weak.

(15) If shallow dopants are used, usually very thin layers are necessary to obtain a continuous space charge zone, since the carrier concentration is then usually above 10.sup.16 cm.sup.−3; for this reason also, several layers are required in order to achieve a sufficiently high insulation effect over a sufficient layer thickness, since the breakdown field strength is limited due to the type of material.

(16) The approach with shallow dopants is already similarly realized in semiconductor laser structures. There, the laser diode structured by etching with an exposed p-n junction with an inverted diode structure, i.e. reversed layer sequence of p- and n-conductive layer as compared to the laser diode structure, is overgrown. This creates a diode that blocks around the laser diode during laser operation in the forward direction, which limits the flow of current to the area in the laser structure. However, this is not comparable with the embodiment according to the invention for the layers mentioned here, since on the one hand a 3-dimensional structured sample is overgrown and on the other hand the layer thicknesses of the inversely grown or later operated p-n structure are usually greater than the space charge zone.

(17) In this example, the target is not a middle position of the Fermi level, but the behavior of a diode operated in the reverse direction. The voltages that can be achieved in this way up until the breakdown would be low due to the limitation of the thickness of these blocking diode layers, but would be limited anyway in a laser structure by the current flow of the laser diode in the forward direction through the active area.

(18) An advantageous development of the semiconductor layer stack in the material system of group III nitrides exemplified below consists in doping one of the following dopants in the first layer (A), alternating with in each case the second dopant in the second layer (B) in group III nitride semiconductors: Iron and carbon or Carbon and a donor or Iron and magnesium or Iron and zinc.

(19) Combinations of these dopants in more than two layers or partly in one of several layers are also possible. In the case of doping with shallow donors such as Si or Ge in GaN or the acceptor Mg in GaN, it can also be advantageous in individual cases to dope them continuously at a low concentration and to only dope the compensating impurity, i.e. a deep acceptor or donor, in individual layer sections, which is covered by the claims according to the invention, since the change in position of the Fermi level in a hypothetical infinitely extended individual layer is critical to the successful implementation, which is also given in these cases.

(20) Donors in group III nitrides can be deep, such as those that can also be achieved with C, or shallow ones, such as Si, Ge or O. However, when using shallow dopants, due to the smaller space charge zone widths, very thin layers or very low dopant concentrations are necessary in order to not produce layers with high residual conductivity, i.e. to obtain an area fully depleted of charge carriers as shown in FIG. 5, where 10 nm thin individual layers are used. The implementation of such thin layers is made more difficult by flow effects and the entrainment of some dopants such as Mg in modern growth processes such as MOVPE.

(21) For the implementation according to the invention or a method for producing a semiconductor layer stack, at least the following steps are advantageous: the provision of a substrate in a system for the deposition of semiconductors the application of a sequence of at least two layers (A, B), which as individual layers each have an energy position of the Fermi level (103) in the semiconductor band gap (104, 105),

(22) $E_F-E_V<\frac{E_G}{2}$  applying to layer (A) and

(23) $E_L-E_F<\frac{E_G}{2}$  applying to layer (B) wherein the thickness (106, 107) of the layers D.sub.A and D.sub.B (A, B) is selected such that a continuous space charge zone region (110)

(24) $W_A+W_B=\sqrt{\frac{2.Math.{\mathrm{.Math.}}_s}{q}\left(\frac{N_A+N_D}{N_A.Math.N_D}\right){\Psi }_{bi}}$  over the layers (A, B) results and D.sub.A≤W.sub.A and D.sub.B≤W.sub.B applies, with W.sub.A and W.sub.B being the space charge zone as well as N.sub.A and N.sub.B the dopant concentrations in layers A and B, ε.sub.S the dielectric constant, q the elementary charge and Ψ.sub.bii the built-in potential difference, which is identical to the energy difference of the Fermi levels.

(25) The choice of the maximum thickness of the layers results from a numerical estimate of the space charge zone widths that are formed, wherein it is advantageous to remain well below the calculated layer thickness, i.e. by at least a factor of two, better five and ideally over 10, since there is then a low band modulation, i.e. a very even band profile.

(26) Below, some exemplary embodiments and a description of the figures are presented using the example of gallium nitride that has already been introduced.

(27) Nowadays, gallium nitride is an important semiconductor for a variety of applications such as LEDs for general lighting but also for power electronics. As a thin-film material, components are usually manufactured commercially using MOVPE.

(28) Due to intrinsic defects and impurities, GaN is usually slightly n-type conductive and rarely has a high impedance, mostly due to process-related carbon impurities. In early work, the acceptors zinc or magnesium were used, which have activation energies >150 meV in order to achieve high impedance, but with a moderate blocking behavior which is above 10.sup.10 cm.sup.−3 at the relatively high hole concentration generated by these dopants. Iron, a frequently used compensator in III-V semiconductors, nowadays is partly used for GaN. However, the energy position of 0.6 eV below the conduction band leads to a relatively high residual electron conductivity, which is disadvantageous in most electronic components. All the more so as unipolar devices in the GaN system often conduct electrons and therefore, electron injection into the insulating layer of such a layer is only moderately well blocked. Alternatively, carbon is used. Intrinsically doped, it results in a deep acceptor in the lower half of the GaN band gap (approx. E.sub.V+0.9 eV) and a donor in the upper half of the band gap. Doped with a precursor such as propane or another hydrocarbon or carbon-containing precursor such as CBr.sub.4 under standard growth conditions, carbon is mainly incorporated as a deep acceptor.

(29) No advantageous properties are reported regarding incorporating carbon as a deep acceptor and, at the same time, as a deep donor, which is not surprising, since setting a suitable acceptor/donor ratio that would ensure a Fermi level as close to the middle of the band gap as possible is not currently possible due to lack of knowledge. In principle, however, such a single dopant that generates two levels is suitable if the ratio of the two levels can be adjusted by the growth conditions. FIG. 2 and FIGS. 3, 4, and 5 show the energy positions of the bands relative to the Fermi level, which is defined here with a value of 0 eV and is represented by the dashed line.

(30) The upper continuous curve in the respective diagram represents the conduction band, the lower one, the valence band.

(31) FIG. 2a) shows the position of the energy bands when doped with Fe; 2 b) when doped with C. In both cases, the Fermi level is relatively close to the conduction or valence band. Therefore, the residual charge carrier concentration is still relatively high.

(32) If the aim is to achieve a high-impedance individual layer, it is possible, for example, to co-dope an individual layer with Fe and C, wherein the ratio of the dopants must take into account the energy position.

(33) An energy position of E.sub.L−0.6 eV for Fe and E.sub.V+0.9 eV results in a ratio of the concentrations of approx. 5,000. This must be met as precisely as possible, but this is also dependent on the exact, but mostly only imprecisely known, energy position in the band gap. For this reason alone it is much easier to dope the dopants in alternating layers, since then the concentration and the exact energy position of the dopants can vary more or be less precisely known. In an n-type semiconductor, the thickness of the Fe-doped layers is thicker than that of the C-doped, since the acceptors only capture electrons and the Fermi level, due to a GaN:C layer that is too thick, is roughly at the energy position of the C in a multilayer stack. Simulations result in a layer thickness of 5 to 20 times that of the GaN:C layer for the GaN:Fe layer, depending on the application. This is shown by way of example in FIG. 3. The Fermi level in this example is 100-200 meV farther away from the valence band than with a single dopant, which further lowers the concentration at the holes, thereby increasing resistivity.

(34) In another embodiment, the combination of the shallow donor Si and the deep acceptor C, the doping of C in concentrations of e.g. 2×10.sup.18 cm.sup.−3 is done only in thin layers if the Si concentration is continuous or present in individual layers, e.g. 1×10.sup.17 cm.sup.−3, which layers are a little thicker than is mathematically necessary to capture all the free electrons generated by Si, i.e. to take up a little more than 1/10 of the volume in this example. This is shown in FIG. 4 for a short stacking sequence. If the C-doped layers are too thick, the Fermi level again lies near the C-type impurity and there is a slight p-type conductivity. If the same is to be accomplished with Mg and Fe, a sufficient n-type background must be present, whether intrinsic or e.g. through continuous doping, in order to successfully combine them, similar to the combination of C and Fe. However, Mg concentrations that are too high or a GaN:Mg layer that is too thick would lead to a noticeable p-type conduction.

(35) In such a case, a deep donor in the upper half of the band gap in combination with or alternating with Mg is more suitable. Ti or the deep C donor could be used here, provided that these can be specifically incorporated. In principle, it is also easily possible to n- and p-dope with shallow impurities in alternation, so that the complete depletion according to the invention takes place via these layers. FIG. 5 shows this for thin alternating layers; here, the Fermi level is on average almost perfectly in the middle of the energy gap.

(36) Combinations of deep and shallow impurities or only shallow ones are interesting insofar as e.g. Si and Mg are mostly available at group III nitride layer deposition systems, and thus at most one deep dopant, instead of two, must be stored. Easier to handle, however, are processes with dopants whose energy levels are so low that they basically deplete the material, i.e. the charge carrier concentration, when using the same, is below 10.sup.14 cm.sup.−3. At charge carrier concentrations below 10.sup.14 cm.sup.−3, the space charge zone width is >1 μm. If dopants such as C or Fe are used in a concentration above the electron concentration in the undoped case, the charge carrier concentration in the GaN normally lies below 10.sup.10 cm.sup.−3, which entails a space charge zone >100 μm. Therefore, there is only slight band bending between the alternatingly doped layers, mostly present towards the more strongly conductive layers lying below and above, if applicable.

(37) FIG. 1a shows a schematic view of an energy-location-diagram of how, starting from an alternating layer structure with the differently doped layers A and B which are here not yet in electrical contact, said layers line up along the electrochemical potential or the Fermi level. Here, 101 is the conduction band E.sub.C, 102 the valence band E.sub.L, 103 the Fermi level E.sub.Dopant, Fermi, which in FIG. 1a should be identical to the energy position of the dopants, provided that the concentration of each of the dopants is so high that the Fermi level is pinned to the energy position of the latter, which is not absolutely necessary for an embodiment according to the invention. This means that even only partial compensation of the background charge carriers and thus a position of the Fermi level that is not identical to the energy position of the dopant can lead to the desired result but is more difficult to implement in a reproducible manner. The thickness of the layers A and B is indicated by 106 and 107 and the respective band gap energy by 104 and 105, which are identical in homoepitaxy but can in principle also differ. If these layers A and B are brought into electrical contact, a potential difference arises which is compensated for by the charge shift and the resulting band bending. FIG. 1b shows this for the case of very thick individual layers, which are so thick that the resultant space charge zones (109) are thinner than half the thickness of the layer pair. In this case, individual space charge zones are created between the differently doped regions, the first three shaded exemplary areas of which are marked with 109. However, the space charge zones do not extend into the adjacent transition; instead, the bands 101 and 102 are shallow outside the space charge region and the Fermi level 108 is pinned at the impurity level. Only within the space charge zone is it located energetically between the two impurity levels. If the layer thickness decreases, then a region completely traversed with space charge zones or a continuous space charge region (110) results, i.e. an electric field modulated by means of the individual layers as shown in FIG. 1c is present over the entire structure, which in FIG. 1d, for even thinner layers, ends in an ever shallower band profile in which the Fermi level 108 assumes an energy position between that of the impurity levels. For reasons of clarity, the decrease in the layer thickness is not graphically displayed or the horizontal axis of the distance is shown on a different scale than in FIG. 1b. Since the charges that are moved here by the potential differences, i.e. the fixed charges of ionized impurities are now only present in a low concentration, the band bending extends over a large area and on average leads to a nearly shallow band profile. In an actual structure, additional band bending results starting from the surface or the doped layers at the edges of the alternating layer structure, which is not taken into account here and usually overlays the entire structure.

(38) By way of example, FIG. 6 shows the comparison of current-voltage characteristic curves a) and the resistivity b) calculated therefrom between an upper metal contact and the conductive substrate of differently doped buffer layers of a GaN based field effect transistor structure on silicon (111). The sample with an Fe-doped GaN buffer shows the highest current, followed by the C-doped buffer, as is expected from the position of the dopants (Fe≈E.sub.L−0.6 eV; C≈E.sub.C+0.9 eV). If the above doping is used and a sample with an approx. 3 μm thick buffer made of alternately 200 nm thick Fe and C-doped GaN layers is prepared, the current is lower, although Fe-doped or C-doped layers alone have a higher current, i.e. the specific resistivity of the individual layers is lower. In the combination according to the invention of alternately successive Fe— and C-doped individual layers, however, they have the lowest current and highest impedance.

(39) In such structures on silicon, Al-containing intermediate layers are introduced into the GaN in order to avoid thermal cracks. Due to the charges that arise at the boundary layers, it is advantageous to dope the GaN with Fe after the Al-containing intermediate layer and before the one with C, since, seen in the direction of growth, there is a potential for an accumulation of holes above the intermediate layer and below the electrons, which can be ideally compensated in this way.

(40) The application according to the invention can be checked most easily, e.g., with methods of mass spectroscopy such as secondary ion mass spectroscopy (SIMS) or on the basis of the defect luminescence in high-resolution processes such as cathodoluminescence in a scanning electron microscope or a scanning transmission electron microscope. Also, dopants and their place of installation can in part be identified with high spatial resolution Raman spectroscopy.

(41) The invention relates to all semiconductors and dopants. In particular, for some semiconductors with a narrow band gap, due to a compensator, the values for the aspired charge carrier concentration are higher than is described here for GaN with E.sub.G 3.4 eV because these semiconductors have higher charge carrier concentrations and intrinsic conductivities. Like other values, they must be adapted in accordance with the band gap energy and the density of states and the resulting intrinsic charge carrier concentrations in order to be able to achieve an inventive effect. Also, in principle it is possible to use more than two dopants and to use them in more than two alternating layers and as part of an inventive embodiment. The frequently present intrinsic conduction n or p>>n.sub.i can also be used in a layer instead of using a dopant in a layer. A doped layer can also overlap into another, i.e. the doping does not have to end abruptly at the nominal end of the layer and an overlap can also be advantageous, depending on the type and energy position of the dopants. Also, the order of the layers A and B specified in the claims is exchangeable and further layers therebetween are possible as long as the condition of a continuous space charge zone region is satisfied. The description of the position of the Fermi level always refers to the case without external voltage applied to the layers.

(42) The invention being thus described, it will be obvious that the same may be varied in many ways. Such variations are not to be regarded as a departure from the spirit and scope of the invention, and all such modifications as would be obvious to one skilled in the art are to be included within the scope of the following claims.