阅读说明：本技术 半导体层堆叠及其制造方法 (Semiconductor layer stack and method for producing the same ) 是由 A·施特里马特 A·达加尔 于 2019-06-20 设计创作，主要内容包括：本发明涉及一种半导体层堆叠、由其组成的构件和构件模块,以及一种制造方法,其中,所述半导体层堆叠的特征在于至少两个层(A,B),所述至少两个层作为单层分别具有半导体带隙(104、105)中如下费米能级(103)的能量位置：对于所述层(A)适用公式(I),对于所述层(B)适用公式(II),其中,E-F是所述费米能级(103)的能量位置,E-V是价带(102)的能量位置,E-L是导带(101)的能量位置,E-L-E-V是所述半导体带隙E-G(104,105)的能量差,其中,如此选择所述层(A,B)的厚度(106,107),使得在所述层(A,B)上存在连贯的空间电荷区区域(110)。(The invention relates to a semiconductor layer stack, a component and a component module consisting thereof, and a production method, wherein the semiconductor layer stack is characterized by at least two layers (A, B) which each have, as a single layer, an energy position in a semiconductor bandgap (104, 105) of the following Fermi level (103): applying formula (I) for the layer (A) and applying formula (II) for the layer (B), wherein E F Is the energy position of the Fermi level (103), E V Is the energy position of the valence band (102), E L Is the energy position of the conduction band (101), E L ‑E V Is the semiconductor band gap E G (104, 105), wherein the thickness (106, 107) of the layers (a, B) is selected such that there is a coherent space charge region (110) on the layers (a, B).)

1. A semiconductor layer stack, characterized in that,

at least two layers (A, B) having as a single layer in the semiconductor band gaps (104, 105) the following energy positions of the Fermi level (103), respectively: for the layer (A) applyAnd for said layer (B) appliesWherein E is_FIs the energy position of the Fermi level (103), E_VIs the energy position of the valence band (102), E_LIs the energy position of the conduction band (101), E_L-E_VIs the semiconductor band gap E_G(104, 105), wherein the thickness (106, 107) of the layers (a, B) is selected such that a coherent space charge region (110) is created on the layers (a, B).

2. The semiconductor layer stack of claim 1, whereinExtrinsic or intrinsic doping is induced in at least two impurity levels at a concentration, said at least two impurity levels having the following energy positions: for the energy position, in one layer (A)And in another layer (B)Such that the fermi level at such an infinitely thick monolayer occupies the same energy position as the impurity level with a tolerance of ± 50meV, wherein the thickness of the layers in the semiconductor layer stack is selected such that a coherent space charge region (110) is constructed over the entire layer stack.

3. The semiconductor layer stack of claim 1 or 2, characterized by the following energy positions of the deep impurity levels generated by the doping: applying in the layer (A) for the energy positionAnd for said energy position in said layer (B)

4. The semiconductor layer stack of claim 1 or 2, characterized in that an average fermi energy location E in the following energy range in the space charge region (110)_F(108)：

5. The semiconductor layer stack according to at least one of the preceding claims, characterized in that doping with an acceptor-type or donor-type dopant.

6. The semiconductor layer stack according to at least one of the preceding claims, characterized in that doping with an acceptor-type and a donor-type dopant.

7. The semiconductor layer stack according to at least one of the preceding claims, characterized in that, in the group III nitride semiconductor, one of the following dopants is alternately doped in the first layer (a) and the second dopant in the second layer (B), respectively:

iron and carbon, or

Carbon and a donor, or

Iron and magnesium, or

Iron and zinc.

8. Semiconductor layer stack according to at least one of the preceding claims, characterized by a sequence of at least two layer groups comprising at least two layers (a, B) having the following fermi level positions: for the Fermi level position, as a single layer in the layer (A)And for said fermi level positions, as a single layer in said layer (B)

9. A building block comprising at least one building block comprising a semiconductor layer stack according to any of the preceding claims.

10. A method for manufacturing a semiconductor layer stack, the method comprising at least the steps of:

providing a substrate in an apparatus for depositing a semiconductor;

applying a sequence of at least two layers (A, B) having as a monolayer in the semiconductor bandgaps (104, 105) the following energy positions of the fermi level (103), respectively: for the layer (A) applyAnd for said layer (B) appliesWherein layer D is selected such that_AAnd D_B(A, B) of a thickness (106, 107) such that a coherent space charge region (110) is obtained on said layer (A, B)And applies to D_A≤W_AAnd D_B≤W_BWherein W is_AAnd W_BIs a space charge region, N_AAnd N_BIs the dopant concentration, ε, in the layers A and B_sIs the dielectric constant, q is the unit charge, Ψ_biIs the same embedded potential difference as the energy difference of the fermi level.

Technical Field

The invention relates to a semiconductor layer stack and a method for producing the same.

Background

The insulating semiconductor layer is not replaceable for electrical insulation and low high-frequency attenuation of the semiconductor component structure. So-called deep impurities are mostly used hereI.e., impurities that ionize only to a low degree (i.e., < 50%) at operating temperature. Doping is usually carried out using deep acceptors (Akzeptor) if the electronic conduction of the semiconductor is present, either intrinsically or via impurities, whereas in the case of hole-intrinsic conduction, doping is carried out using deep donors (donators). Since the electron injection into the separating layer is correspondingly well prevented in the first case and the hole injection in the second case, the insulating properties are improved by the combined simultaneous doping of deep acceptors and deep donors (e.g. deep acceptors Fe and deep donors Ti) in InP [ t.wolf, t.zinke, a.krost, h.scheffller, h.ullrich, p.harde and d.biberg, j.appl.phys.75,3870(1994)]. However, usually only one dopant (Dotand) is doped, since this is sufficient in most cases for achieving an insulating effect and is also easier to control in terms of process technology. Ideally, a deep impurity is present near the center of the semiconductor band gap and at a concentration such that it is able to capture all free charge carriers and thus the fermi level is located or fixed at the energy location of the impurity, even with charge carrier injection. In the ideal case, the pair of free charge carrier concentrations of electrons n and holes pCorresponding to the intrinsic charge carrier concentration n_iI.e. n-p-n_i。

In many semiconductors, including those with large energy gaps, such as group III nitrides, there are, in principle, many dopants that can build deep impurities and high resistance materials, but their location is generally not near the center of the semiconductor band gap. Therefore, even in the case of a deep impurity (N) having a sufficient concentration N for compensation>n, p), i.e. at a concentration higher than the concentration n or p of electrons or holes without the presence of a compensating agent, there is a residual conductivity despite the Fermi level being fixed theretoAnd n or p is much higher than the intrinsic charge carrier concentration n_i。

Even if the charge carrier concentration is less than 10¹⁰cm^-3This is also considered high in some semiconductors, since for example very high breakdown field strengths of more than 3MV/cm may occur in semiconductors such as GaN, and low leakage currents at high applied voltages or field strengths are clearly disadvantageous. Furthermore, this is because the performance achieved locally at high voltages even at low leakage currents may lead to heating and ultimately to thermally induced breakdown. The stress therefore requires as low a residual conductivity or charge carrier concentration as possible, which is as close as possible to the intrinsic charge carrier concentration. In GaN, for example, the carbon acceptor is about 0.9eV higher than the valence band, while the iron acceptor that can be used instead is about 0.6eV lower than the conduction band. However, at room temperature, when the bandgap of GaN is about 3.4eV, the ideal location for the deep impurity is about 1.7 eV. Thus for GaN: c, free hole concentration of about 1X 10⁶cm^-3For GaN Fe, the free electron concentration is about 3X 10⁶cm^-3Well above about 10^-9cm^-3The intrinsic charge carrier concentration of (a). As a result, the layer resistance when using these dopants is more than 10 orders of magnitude lower than what theoretically could be achieved with ideal dopants (which have energy positions near the center of the bandgap). Since such dopants, for example in GaN, are hitherto unknown or cannot be used in a suitable form for layer manufactureThe performance capabilities of these layers are limited and limit the field of application of the components made from these layers.

Disclosure of Invention

The object is now to achieve improved layer insulation. This object is achieved by means of a semiconductor layer stack according to claim 1, a component module according to claim 9 and a method according to claim 10.

A semiconductor layer stack is proposed, characterized in that at least two layers (A, B) are provided as a single layer, each having the following energy positions of a Fermi level (103) in a semiconductor bandgap (104, 105): for the layer (A) applyFor the layer (B) applyWherein E is_FIs the energy position of the Fermi level (103), E_VIs the energy position of the valence band (102), E_LIs the energy position of the conduction band (101), E_L-E_VIs a semiconductor band gap E_G(104, 105), wherein the thickness (106, 107) of the layer (A, B) is selected such that there is a coherent layer (A, B)A space charge region (110). The layers can be separated directly one after the other or by further layers.

According to the present invention, the problem of locating the fermi level near the center of the energy gap of the semiconductor is solved.

A coherent space charge region is achieved if successive segments of the electric field falling or having different potential courses are connected to one another over the entire region. The width of such space charge region between two layers depends mainly on how high their doping levels (i.e. how many ionizable acceptors and donors are contained), the embedded (eingebaut) potential difference Ψ_bi(which corresponds to the difference in fermi levels in the monolayer) and the dielectric constant ε_SThe value of (c). The width can be calculated. For having shallow acceptor N_AAnd shallow donor N_DP-n junction of (1), in the p region W_pNeutralizing n region W_nThe result of estimating the space charge region width in (1) is

Wherein q is the unit charge (Elementarladung).

The calculations are somewhat complex for deep impurities and multiple dopants or defects that otherwise result in residual conductivity. The value of the space charge region width can in principle be well estimated from the difference in the fermi level positions and the concentration of ionized acceptors and donors in the space charge region, which can be estimated from its energy positions. In this case, it is desirable to achieve values far above the layer thickness of the individual layers in order to then achieve the fermi level position between the initial dopants on average. The dopant may be an acceptor or donor depending on the conductivity of the starting material. Alternatively, in the case of an undoped n-conducting material, it is also possible to dope deep acceptors in the first layer with a concentration higher than the electron concentration, and to dope acceptors and deep donors in the second layer with a higher concentration.

Considered separately, each layer is depleted of charge carriers, whereas the position of the fermi level is close to the position of the deep acceptor or donor (which in rare cases is located near the centre of the bandgap) and therefore a relatively high residual conductivity remains for the monolayer. However, by the alternating growth of these layers, a space charge region is generated which enables the fermi level position to be in between the respective impurity levels and, therefore, enables a lower concentration of charge carriers in the entire layer.

In principle, the embedding of impurities and thus the position of the fermi level can be controlled within a confined framework by the manufacturing conditions. For directional doping it is desirable to use dopants provided from the starting material (intrinsic) or by a special source (extrinsic) during growth.

A particularly advantageous embodiment is to use at least two impurity levelsTo a concentration that causes extrinsic or intrinsic doping, the at least two impurity levels having respectively the following energy positions: in one layer (A)And in another layer (B)Such that the fermi level occupies the same energy position as the impurity level in such an infinitely thick monolayer with a tolerance of ± 50meV, wherein the thickness of the layers in the semiconductor layer stack is selected such that a coherent space charge region (110) is formed over the entire layer stack.

It is advantageous, in particular, not only for the two layers to be grown one after the other, but also for them to alternate alternately, wherein further layers with further dopants or conductivities can also be integrated into the layer stack, as long as the space charge region extends over the entire stack. Typically, space charge region regions also exist when only one deep impurity is used as a compensator within the structure of the component. However, subsequent energy position insertion (pinen) of the fermi level can be disadvantageous. By alternating layers and the fermi level positions which are thereby changed concomitantly, potential fluctuations and therefore space charge regions and electric fields also occur between the individual layers when a single layer is observed. This results in an even further reduction in the charge carrier concentration and approaches n_iThis means a higher layer resistance. This in turn enables the concentration of the dopants used to be reduced, which can have a positive effect on the switching properties (Schaltverhalten) of the component layers situated above it.

The doping thus results in an effective fermi-energy position which, in addition to the fermi-energy position in the individual layers, also depends on the concentration of the dopant and on the thickness of the individual layers, wherein the resulting position can be optimally set by means of the thickness parameters.

Preferably, the semiconductor layer stack is connected toThe energy positions of the deep impurity levels resulting from the over-doping are as follows: in the layer (A) is suitableIn the layer (B)I.e., the average fermi level in the region is below or above the mid-range of the bandgap energy, where the mid-range amounts to 40% of the bandgap energy. In many semiconductors, the value of the available impurities as individual dopants exceeds 50%, which leads to unsatisfactory compensation, as is the case, for example, in GaN: Fe. If the energy level generated is close to the centre of the bandgap energy, i.e. in the range of 40% defined above, satisfactory results can generally be achieved in terms of layer insulation even with this sole dopant.

However, the method according to the invention can further improve this and can therefore also be used meaningfully in individual cases. Background doping of the layer by intrinsic defects or unintentional impurities is also considered to be a dopant and is in particular an intrinsic dopant here. This can be used alternately with intentionally doped layers in a single layer. However, due to the usually slightly fluctuating background doping of the semiconductor layer, it is always preferable to use intentionally doped layers. Intrinsic dopants may be, for example, carbon or oxygen, which may be derived from the starting material or from a carrier gas, depending on the production processBy suitable selection of the growth parameters, carbon intercalation (einbauen) from these dopants, in particular from the alkyl groups of the metalorganic used, can be reproducibly carried out by means of metal-organic vapor phase epitaxy (MOVPE).

Since defects (in particular the breakdown strength and the deviation of the layer resistance, here exemplified by GaN) also use lower limit values as limiting factors, a sufficient center position of the fermi level caused by deep impurities is also of interest at a certain distance from the theoretically most favorable position.

Seeking an energy range in a space charge region (110) in a semiconductor layer stack Average fermi energy position E in_F(108) I.e. the average fermi energy position is in the following range above or below the average energy of the band gap, respectively: said ranges are respectively up to 20% of the band gap energy. For simplicity, the band gap center is defined as n-p-n_iThe ideal fermi level position of the semi-insulating semiconductor. Average position means that the fermi level lies, on average, within this range over the thickness of the buffer region according to the invention, i.e. also in a range above or below this range is allowed, as may occur in the alternating doping according to the invention. The fermi level is as always in this range as possible, which is ideally optimized by choosing a sufficiently thin layer, a single layer of the layer stack doped with the dopant. To such an extent that 1/4 should be dropped in the layer pair due to space charge regions, preferably below the possible valence or conduction band energy change for an infinitely extended layer pair.

In principle, simultaneous doping of two dopants in one layer brings the advantages according to the invention, however here a highly precise control of the respective concentrations is required in order to ideally locate the fermi level. However, if individual layers with alternating doping are involved, the following effects are used here: a space charge region with a slightly curved band is formed between the two doped regions and depends very sensitively on the respective dopant concentration, provided that the fermi level in the monolayer is fixed at the impurity level induced by the dopant. In this case, the band variation process (Bandverlauf) can also be calculated and reproducibly produced.

Due to the low carrier concentration when compensating for excess carriers, the space charge region will typically extend over a region of a few microns (or less if the differently doped layers are chosen to be sufficiently thin), thus resulting in an almost flat band variation process with a slight modulation of the energy of the respective band in each region, as shown in fig. 1 d.

The desired layer thickness depends here on the intrinsic background doping, the type of impurity (i.e. acceptor or donor) and the energy position. In individual cases, the optimum doping must be found for this either simulatively or experimentally. In principle, in semiconductors with n-type background doping, the charge carriers are mostly already trapped by the acceptors located in the upper half of the energy gap. Charge carriers still remaining in the conduction band are then trapped by a second acceptor in the lower half of the band gap. The resulting space charge region is then so wide that the fermi level is located close to the lower acceptor almost throughout the layer stack, and therefore the layers must be very thin so that the fermi level does not drop too strongly and on average the position at which the fermi level is achieved is close to the centre of the bandgap, i.e. without inducing noticeable hole conduction. However, even if the optimum thickness is not selected, as shown in fig. 6, since it is possible to achieve strong depletion of charge carriers locally, improvement of layer insulation can be achieved. Furthermore, even in the absence of a voltage applied across the layer stack, the position of the fermi level has been dependent on how the adjoining layers are doped.

For example, if one chooses to pass the shallow donor lower (i.e., at about 10)¹⁶-10¹⁷cm^-3In the range of (b) but targeted n-doping and compensating this with deep acceptors in the lower half of the band gap, it is advantageous to dope deep acceptors in high concentrations (much higher than the electron concentration), but only in thin layers for this. This results in a complete depletion of electrons in the semiconductor, which has only a low hole concentration through a thin layer with deep acceptors, unlike in the case of a layer doped only with deep acceptors. These thin layers, which are highly doped with deep acceptors, must be placed so densely that the resulting space charge region regions overlap. In this case, it is also necessary to dope the deep acceptors at the beginning and end of the intentional insulation layer stack, respectively, in order to obtain a completely insulating and non-conductive edge region. Such a layer is advantageous for the switching behavior of the component, sinceAt varying applied voltages, re-emission (Reemission) of charge carriers from deep impurities is minimized by the reduced amount of deep impurities and at the same time the reduced residual conductivity.

The doping according to the invention in the semiconductor layer stack can be a doping with an acceptor-type or donor-type dopant or else a doping with an acceptor-type and donor-type dopant. Although doping with only acceptor type or only donor type can be achieved and sometimes also meaningful, theoretically a combination of donor and acceptor is desirable because they better compensate electron and hole injection, as known for Fe and Ti compensators in InP, where they are very close in energy to the center of the energy gap and therefore do not require adjustment of the fermi level according to the invention by a combination of two dopants. If acceptor and donor are used, it is not easy to dope them in the alternating layers due to the background doping of the electrons or holes that are usually present, since deep donors do not trap or compensate electrons, whereas deep acceptors do not trap or compensate holes. In the case of, for example, n-type background doping, this can be solved in the following way: the donor is co-doped with the shallow p-type dopant at a concentration higher than the electron concentration.

Alternatively, one of the upper band gap halves is used as a donor and one of the lower band gap halves is used as an acceptor. In GaN, the latter can achieve, for example, C as a deep acceptor in the first layer and a deep donor without p co-doping in the upper half of the bandgap in the second layer. If a space charge region is constructed on this region, then the slight p-type conduction induced by the acceptor is compensated by the donor in the second layer, albeit with two energy positions both well outside the central region of the energy bandgap. A structure having a deep acceptor in layer a and a deep donor in layer B is desirable. In such a combination, it is preferable to use an acceptor in the lower half of the bandgap and a donor in the upper half of the bandgap.

The structure or semiconductor layer stack according to the invention consists of at least two layers, i.e. a sequence of at least two layer groups (schichpaket), which comprises a structure having the following fermi levelAt least two layers (a, B): for the Fermi level, as a single layer in the layer (A)As a single layer in the layer (B)It is advantageous, among others, for the layers to be alternated a plurality of times, i.e. for ABABAB, ABABA or BABAB to be grown, for example ABCBA or ACBADB, etc., layer sequences are also possible, in which C and D are any intermediate layers, but do not allow to prevent the formation of coherent space-charge-region regions (110) due to their thickness and doping.

According to the invention, a component module is proposed, which comprises at least one component which comprises a semiconductor layer stack according to the invention.

In principle, it is also possible for the structures according to the invention to have only shallow dopants, or to combine shallow and deep donors and/or acceptors. I.e., deep and shallow donors or, conversely, only shallow donors and shallow acceptors as shown in fig. 5. In this case, a multilayer (mehrfachcschicht) is required for the insulating effect, since a p-n junction, i.e. a diode structure, is produced in only one pair of layers a and B. At the deep fermi level according to the invention, a p-n structure is also produced in principle, however, these layers are then already highly resistive, so that the current through these layers is greatly reduced and the diode characteristic curve is only very weakly pronounced.

If shallow dopants are used, mostly very thin layers are required to achieve a coherent space charge region, since the charge carrier concentration is typically higher than 10¹⁶cm^-3For this reason too, a plurality of layers is required to achieve a sufficiently high insulation effect at a sufficient layer thickness, since the breakdown field strength is limited by the material.

Schemes with shallow dopants have been implemented in a similar manner in semiconductor laser structures. There, the laser diode is overgrown with an inverted diode structure (i.e. an inverted layer sequence of p and n conducting layers compared to the laser diode structure) which is constructed by etching and has an exposed p-n junction. Thus, in laser operation, an off diode is generated around the laser diode operating in the on direction, which restricts the flow of current to regions in the laser structure. This cannot be compared with the embodiment according to the invention for the layers mentioned here, however, since the 3-dimensionally structured samples on the one hand overgrow and the layer thickness of the inversely grown or later-on p-n structures on the other hand is generally greater than the space charge region.

Here, in this example, the target is not the center position of the fermi level, but the performance of the diode operating in the off direction. By limiting the thickness of these diode layers which are switched off, the voltage which can be achieved thereby is low up to the breakdown voltage, but in laser structures this is limited anyway by the current of the laser diode in the conducting direction through the active region.

The advantageous further development of the semiconductor layer stack in the group III nitride material system listed below by way of example consists in that, in the group III nitride semiconductor, one of the following dopants is alternately doped in each case in the first layer (a) and in each case in the second layer (B):

iron and carbon, or

Carbon and a donor, or

Iron and magnesium, or

Iron and zinc.

Here, combinations of these dopants above more than two layers or partially in one of the layers are also possible. In the case of doping with shallow donors (for example Si or Ge in GaN) or acceptor Mg in GaN, it is likewise advantageous in individual cases to dope them continuously at low concentrations and to dope only the individual layer sections with compensating impurities (i.e. deep acceptors or donors), which is also covered by the claims according to the invention, since the change in the position of the fermi level in the assumed, infinitely extending monolayer is decisive for successful implementation, which is also given in this case.

Donors in group III nitrides may be deep, e.g. can also be achieved by C, andmay be shallow, such as Si, Ge or O. However, when using shallow dopants, a very thin layer or a very low dopant concentration is necessary in order not to produce a layer with a high residual conductivity due to the small space charge region width, i.e. to obtain a region completely depleted of charge carriers, as shown in fig. 5, in which a thin monolayer of 10nm is used. Due to the effect of flowAnd the lingering of certain dopants (e.g., Mg) in modern growth processes (e.g., MOVPE) (Verschleppung), it becomes more difficult to achieve such thin layers.

For the implementation according to the invention or due to the method for producing a semiconductor layer stack, at least the following steps are advantageous:

providing a substrate in an apparatus for depositing a semiconductor;

applying a sequence of at least two layers (A, B), each layer having as a monolayer the following energy positions of the fermi level (103) in the semiconductor bandgap (104, 105), respectively: for the layer (A) applyFor layer (B)

Wherein layer D is selected in this way_AAnd D_B(A, B) such that a coherent space charge region (110) is obtained on the layer (A, B)And applies to D_A≤W_AAnd D_B≤W_BWherein W is_AAnd W_BIs a space charge region, N_AAnd N_BIs the dopant concentration in layers A and B, ε_sIs the dielectric constant, q is the unit charge, Ψ_biIs an embedded potential difference that is the same as the energy difference of the fermi level.

The choice of the maximum thickness of the layer results from a numerical estimate of the width of the structured space charge region, wherein it is advantageous to keep the layer thickness significantly (i.e. at least twice, preferably five times, ideally more than 10 times) below the calculated layer thickness, since then a low band modulation, i.e. a very uniform band variation process, is provided.

The following description of some embodiments and figures is shown in the gallium nitride example that has been introduced.

Detailed Description

Gallium nitride is an important semiconductor in various applications today, such as LEDs for general illumination, and also for power electronics. Components are often manufactured commercially by means of MOVPE as film material.

Due to intrinsic defects and impurities, GaN is typically slightly n-type conducting and rarely highly resistive, mainly due to process-related carbon impurities. Early work used acceptors of zinc or magnesium with activation energies greater than 150meV to achieve high resistance, but with moderate cut-off performance due to the higher than 10 generated by these dopants¹⁰cm^-3Relatively high hole concentration. Iron is a commonly used compensator in III-V semiconductors and is now partially used for GaN. However, energy positions below the conduction band of 0.6eV lead to a relatively high residual electron conductivity, which is disadvantageous in most electronic components. This is especially the case since the unipolar components in GaN systems are usually electron conducting and therefore such layers only moderately well block electron injection into the insulating layer. Carbon is used instead. Intrinsic doping results in the lower half of the GaN bandgap (about E)_v+0.9eV) and donors in the upper half of the bandgap. Under standard growth conditions, in the case of doped precursors (Precursor), such as propane, or other precursors containing hydrocarbons or carbon, such as CBr4, the carbon is mainly embedded as a deep acceptor.

No favorable properties are reported with respect to the intercalating carbon as a deep acceptor and at the same time as a deep donor, which is not surprising since, due to lack of knowledge, it is not presently possible to set a suitable acceptor/donor ratio to ensure that the fermi level is as close as possible to the centre of the bandgap. In principle, however, such single doping (Einzeldotand) to produce the two energy levels is suitable if the ratio of the two energy levels can be set by growth conditions. The energy position of the band with respect to the fermi level, which is here by definition the value 0eV and is represented by a dashed line, is shown in fig. 2 and fig. 3, 4 and 5.

The upper continuous curve in the corresponding figure represents the conduction band and the lower one the valence band.

Fig. 2a) shows the position of the energy band when doped with Fe, and 2b) shows the position of the energy band when doped with C. In both cases, the fermi level is relatively close to the conduction or valence band. The remaining charge carrier concentration is still relatively high.

If a high resistance monolayer is to be obtained, the monolayer can be co-doped with Fe and C, for example, wherein the proportion of dopants has to take into account the energy position. E for Fe_L0.6eV and E_VAn energy position of +0.9eV, giving a concentration ratio of about 5000. This must be met as precisely as possible, but at the same time, it depends on the exact, but usually only inaccurately known, location of the energy in the bandgap. It is therefore only much easier to dope the dopants in the alternating layers, since then the concentration and the exact energy position of the dopants allow for more intense variation or do not have to be known exactly. Thus, in an n-type semiconductor, the thickness of the Fe doped layer is thicker than that of the C doped layer because the acceptor only captures electrons, and the fermi level is increased due to the excessively thick GaN: c layer is located approximately at the energy location of C in the multilayer stack. According to the application condition, the GaN is obtained through simulation: the layer thickness of the Fe layer is GaN: 5 to 20 times of layer C. This is shown schematically in fig. 3. In this example, the fermi level is further from the valence band by 100-200meV than with a single dopant, which further reduces the concentration of holes, thereby increasing resistance.

In another embodiment-combination of shallow donor Si and deep acceptor C, the Si concentration present in a continuous or in a single layer is, for example, 1X 10¹⁷cm^-3In the case of (2X 10), for example, the C concentration¹⁸cm^-3Is doped only in a thin layer, which is thicker than necessary for pure calculation, in order to trap all the free electrons generated by the Si, i.e. in this example occupy more than1/10 of volume. This is shown in fig. 4 in a short stacking sequence. If the layer doped with C is too thick, the fermi level is again close to the C impurity and a slight p conductivity may occur. If similarly implemented with Mg and Fe, there must be sufficient n-type background, whether it is intrinsic or can be successfully combined, for example by successive doping, like the combination of C and Fe, only too high Mg concentration or too thick GaN: the Mg layer results in significant p-type conductivity.

In this case, it is more appropriate that the deep donors in the upper half of the bandgap combine or alternate with Mg. Ti or deep C donors can be used here, provided they can be embedded in a targeted manner. In principle, it is also possible to alternate slight n-doping and p-doping by means of shallow impurities, so that a complete depletion according to the invention is achieved by these layers. In fig. 5, thin alternating layers are shown where the fermi level is almost perfectly centered in the gap.

Since Si and Mg are usually present in e.g. a group III nitride layer deposition apparatus and therefore at most one deep dopant has to be retained instead of two, a combination of deep and shallow impurities or only shallow impurities are of interest. However, processes with dopants whose energy levels are so deep that they in principle deplete the material, i.e. with a carrier concentration of less than 10, are easier to handle¹⁴cm^-3. Thus, the carrier concentration is less than 10¹⁴cm^-3Giving a space charge region width of more than 1 μm. If a dopant such as C or Fe is used in a concentration higher than the electron concentration without doping, the carrier concentration in GaN is usually less than 10¹⁰cm^-3This will bring a space charge region of more than 100 μm. Thus, the band bending between the alternately doped layers is only slightly pronounced and mainly bends towards the stronger conductive layers above and below it, if present.

In fig. 1a, an energy position diagram is schematically shown, showing how, starting from an alternating layer structure with differently doped layers a and B (which are not yet in electrical contact here), they are arranged along the electrochemical potential or fermi level. Where 101 is the conduction band E_C，102Is valence band E_LAnd 103 is the Fermi level E_Dopant,FermiThe fermi level in fig. 1a should be the same as the energy position of the dopant, which is not necessarily necessary for the embodiment according to the invention if the concentration of the dopant is respectively so high that the fermi level is fixed at the energy position of the dopant. That is, compensation of the only partial presence of background charge carriers and the location of the fermi level (which is not the same as the energy location of the dopant) can lead to desirable results, but is difficult to achieve reproducibly. The thicknesses of the layers a and B are indicated with 106 and 107 and the corresponding band gap energies are indicated with 104 and 105, which are the same homoepitaxy, but can in principle also be different. If these layers a and B are brought into electrical contact, a potential difference is generated which is compensated by the charge offset and the resulting band bending.

In fig. 1b, a very thick monolayer is shown, which is so thick that the resulting space charge region (109) is thinner than half the thickness of the layer pair. In this case, separate space charge regions are created between the differently doped regions, the areas marked 109 for the first three exemplarily shaded areas. However, the space charge region does not extend into the adjacent transition region, but outside the space charge region, the bands 101 and 102 are flat, and the fermi level 108 is fixedly located at the impurity level. Only in the space charge region is its energy between the two impurity levels. If the layer thickness is reduced, a region is created which is completely penetrated by the space charge region or the coherent space charge region (110), i.e. there is an electric field modulated by the individual layers over the entire structure, as shown in fig. 1c and 1d, for thinner layer passes to an always flat band transition in which the fermi level 108 is at an energy position between the impurity levels. For the sake of clarity, the reduction in layer thickness is not graphically realized here, or the horizontal axis of the spacing is shown on a different scale than in fig. 1 b. Since the charges moved by the potential difference or the fixed-position charges of the ionized impurities are present here only in low concentrations, the band curvature extends over a large area and results in an average almost flat band course. In practical structures, additional band bending is obtained starting from the surface or doped layer at the edge of the alternating layer structure, which band bending is not taken into account here, which mostly covers the entire structure.

Fig. 6 shows a comparison of the current-voltage characteristic curve a) and the resistance b) calculated therefrom between the upper metal contact of the differently doped buffer layer of the GaN-based field effect transistor structure on silicon (111) and the conductive substrate. Also as from the position of the dopant (Fe ≈ E)_L-0.6eV；C≈E_C+0.9eV), where the sample with the Fe doped GaN buffer layer shows the highest current, followed by the C doped buffer layer. If this doping is now carried out and samples with buffer layers of about 3 μm thickness are produced, which consist of alternating 200nm thick layers of Fe and C-doped GaN, the current is lower, although the Fe-doped or C-doped layers alone have a higher current, i.e. the resistivity of the individual layers is low (spezifische widersland). However, in the combination of alternating successive Fe and C doped monolayers according to the invention, it has the lowest current and the highest resistance. In this structure on top of silicon, an aluminum-containing intermediate layer is introduced into GaN to avoid thermally defined cracks. Here, it is advantageous to dope the GaN with Fe after the Al-containing intermediate layer, and before this with C, because, as seen in the growth direction, a potential accumulation of holes occurs above the intermediate layer and an accumulation of electrons occurs below it, which can be compensated for ideally, on account of the charges generated at the boundary surfaces.

The application according to the invention can be checked most simply, for example by means of mass spectrometry methods such as Secondary Ion Mass Spectrometry (SIMS), or on the basis of defect luminescence in high-resolution methods, such as cathodoluminescence in scanning electron microscopes or scanning transmission electron microscopes. It is also sometimes possible to identify the dopant and its embedded position by means of high-position-resolved raman spectroscopy.

The present invention relates to all semiconductors and dopants. Especially for some semiconductors with a smaller band gap, the value of the charge carrier concentration sought by the compensator is higher than for the GaN (E) described here_G3.4eV) because these semiconductors have a high charge carrier concentration and intrinsic conductivity. Like other values of oneAs such, it must also be matched in relation to the band gap energy and the density of states and thus the intrinsic charge carrier concentration in order to be able to achieve the effect according to the invention. In principle, more than two dopants and dopants in more than two alternating layers are also possible and are part of an embodiment according to the invention. Instead of dopants in the layer, the intrinsic conductivity n or p > ni that is usually present in the layer can be used. One doped layer may also overlap into the other, i.e. the doping does not have to terminate abruptly at the nominal end of the layer and can overlap, which may also be advantageous depending on the type and energy position of the dopant. The order of the layers a and B mentioned in the claims is interchangeable and there may be further layers between them, as long as the condition of a coherent space charge region area is fulfilled. The description of the position of the fermi level always refers to the case where no external voltage is applied across the layer.