JPH04245649A - Field effect transistor - Google Patents

Abstract

PURPOSE:To obtain a high output FET excellent in high frequency characteristics. CONSTITUTION:On a GaAs semiconductor substrate 21, the following are formed in order; a first semiconductor layer 22 composed of almost undoped GaAs, a first channel layer 23 composed of high concentration thin-layered N<+> type GaAs, a second semiconductor layer 24 composed of almost undoped GaAs, a second channel layer 25 similar to the first channel layer 23, and a third semiconductor layer 26 composed of almost undoped GaAs. Further on the third semiconductor layer 26, a fourth semiconductor layer 27 composed of almost undoped AlxGa1-xAs (0<x<=0.3) is formed. The fourth semiconductor layer 27 forms a hetero junction together with the third semiconductor layer 26, and is in Shottky contact with a gate electrode 32.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the structure of a field effect transistor (FET) which requires ultra-high speed operation.

0002

2. Description of the Related Art Conventionally, as this type of ultra-high-speed device, there is, for example, a HEMT (high electron mobility transistor) having a structure shown in FIG. An undoped GaAs layer 2 is formed on a GaAs semiconductor substrate 1, and an n-AlGaAs layer 3 in which donors are selectively added to AlGaAs is formed on this undoped GaAs layer 2. Furthermore, on this n-AlGaAs layer 3, n+ -Ga
An As layer 4 is formed, and a gate electrode 5 is formed in Schottky contact with the n-AlGaAs layer 3 exposed in a recess formed at the center. Also, n+-GaAs
Ohmic electrodes 6 and 7 are formed on layer 4.

[0003] Other ultra-high-speed devices include:
For example, DMT (Doped) with the structure shown in FIG.
-channel hetero MIS-FET) has been developed. An undoped GaAs layer 12 is formed on the GaAs semiconductor substrate 11.
On the s layer 12 is an n+ -GaAs layer 13 which becomes a channel.
is formed. Furthermore, this n+ -GaAs layer 1
3, an undoped AlGaAs layer 14, n+ -Ga
An As layer 15 is formed. Further, a gate electrode 16 is formed in Schottky contact with the undoped AlGaAs layer 14 exposed in the recess, and the n+ -GaAs layer 1
Ohmic electrodes 17 and 18 are formed on 5.

[0004] In addition to this, Japanese Patent Application Laid-Open No. 64-82677
There is also an ultra-high speed device with a two-layer planar doped structure disclosed in the publication. In this device, the impurity is 2
Two planar doped layers doped in a dimensional plane are provided with an interval within the mean free path of electrons. By using these planar doped layers for the channel layer, the speed of the device is increased.

[0005]

The conventional HEMT shown in FIG. 6 has an undoped GaAs layer 2 and an n-AlG layer.
The two-dimensional electron gas layer 8 generated at the heterojunction interface with the aAs layer 3 is used as a channel. The maximum current density of HEMT is determined by the upper limit of this two-dimensional electron gas concentration, but since the channel layer is two-dimensional, there is a limit to increasing the electron gas concentration. For this reason, it has not been possible to obtain a high frequency device with sufficiently high output.

On the other hand, the conventional DMT shown in FIG.
Since the S layer 13 is used, it is possible to obtain a sufficiently high output. Also, AlGaA in this upper layer
Since the s-layer 14 is undoped, the Schottky breakdown voltage is improved. However, in this DMT, since the channel layer contains a large amount of impurities,
The speed of electrons traveling through the channel layer is lower than in HEMT. As a result, the high frequency operating performance of DMT is
It was inferior to T.

[0007] Furthermore, in systems using AlGaAs/GaAs heterojunctions such as HEMTs and DMTs, real-space transition occurs in which electrons traveling in the GaAs layer, which becomes a channel, transition to the AlGaAs layer above the GaAs layer. may occur. For example, in the conventional HEMT mentioned above, this can be explained as follows. n-Al
An energy band shown in FIG. 8 is formed at the heterojunction between the GaAs layer 3 and the undoped GaAs 2, and two-dimensional electron gas is accumulated in the hatched area shown. However, when a high electric field is applied between the drain and source and the energy of the two-dimensional electron gas increases, the electrons in the two-dimensional electron gas become n
- Transition to the AlGaAs layer 3 side as shown by the arrow.

Generally, a high electric field is applied between the drain and source during operation, and in the AlGaAs layer, GaAs
Since the electron transport characteristics are inferior to those of the layer, when this real space transition occurs, the high frequency characteristics of the FET deteriorate.

Furthermore, in the conventional FET having the two-layer planar-doped structure described above, a two-dimensional planar thin planar-doped layer is used for the channel layer, and the thickness of the planar-doped layer is about 5 to 6 angstroms. It has a thickness of several layers. Therefore, the amount of impurity added to this doped layer is at most about 1×10 13 /cm 2 . For this reason, the electron concentration in the channel layer is limited even if two planar doped layers are provided, and like the conventional HEMT described above, although it is possible to increase the speed of the device, it is hindered from increasing its output. was.

0010

[Means for Solving the Problems] The present invention has been made to solve the above problems, and provides a first semiconductor layer made of a material with excellent electron transport properties that does not contain any impurities or contains impurities at a low concentration. a first channel layer made of thin GaAs containing a high concentration of n-type impurities and having a crystal structure that is approximately lattice-matched to the first semiconductor layer; a second semiconductor layer made of a material with excellent electron transport properties that does not contain any impurities or contains a low concentration of impurities, and has a crystal structure that matches the second semiconductor layer; A second channel layer made of GaAs thinly formed with a high concentration of impurities, and a crystal structure that is almost lattice-matched to the second channel layer, with electron transport properties that do not contain any impurities or contain impurities at a low concentration. A third semiconductor layer made of an excellent material, and AlX Ga1-X A that forms a heterojunction with the third semiconductor layer and contains no impurities or a low concentration of impurities that make Schottky contact with the gate electrode.
s (0<X≦0.3).

[0011]

[Operation] When a low electric field is applied between the drain and source, electrons generated in the first and second channel layers containing high impurity concentrations contain no impurities located between these channel layers. The probability of the second semiconductor layer having excellent electron transport properties increases the probability that the second semiconductor layer does not contain or contains a low concentration of electrons.

When a high electric field is applied between the drain and source, the electrons traveling in the channel gain energy and the first
The electrons then jump out to the first and third semiconductor layers, which sandwich the second channel layers and have excellent electron transport properties.

Furthermore, since each of the first and second channel layers has a certain thickness, they can contain impurities at a high concentration, and the channel is formed by a large amount of electrons.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows the structure of an FET according to an embodiment of the present invention, and its manufacturing method is shown in the cross-sectional views of FIG. 2. This manufacturing method will be explained below.

First, a semi-insulating GaAs semiconductor substrate 21
Above, a first semiconductor layer 22, a first channel layer 23, a second semiconductor layer 24, a second channel layer 25, a third semiconductor layer 26, a fourth semiconductor layer 27, and a contact layer 2.
8 are sequentially epitaxially grown (see FIG. 2(a)). This epitaxial growth is performed by MBE (molecular beam epitaxy) or OMVPE (organic metal vapor phase epitaxy).
It is done by law.

The first semiconductor layer 22 on the semiconductor substrate 21 is made of almost undoped p-type GaAs which has better electron transport characteristics than the channel layers 23 and 25, and has a thickness of 1 μm. The first channel layer 23 and the second channel layer 25 are each made of n+ type GaAs doped with Si at a concentration of 2×10 18 /cm 3 , and each has a thickness of 100 angstroms. The second semiconductor layer 24 sandwiched between each channel layer 23 and 25 is
It is made of nearly undoped n-type GaAs, which has better electron transport characteristics than each of the channel layers 23 and 25, and has a thickness of 100 angstroms. second channel layer 2
The third semiconductor layer 26 on top 5 is made of the same material as this second semiconductor layer and has a thickness of 200 angstroms. The fourth semiconductor layer 27 forming a heterojunction with the third semiconductor layer 26 has an Al composition ratio X of more than 0 and less than or equal to 0.3 (
0<X≦0.3) almost undoped n-type A
It is made of lx Ga1-x As and has a thickness of 200 angstroms. The contact layer 28 on the fourth semiconductor layer 27 is made of n+ type GaAs doped with donors at a concentration of 4×10 18 /cm 3 and has a thickness of 5
00 angstroms.

Here, the carrier concentration and thickness of each channel layer 23 and 25 are set to be sufficient to form a quantum well, which will be described later. Furthermore, since the electrons in each of these channel layers 23 and 25 have energy, the electrons in each of these channel layers 23 and 25 have energy.
, 25 is present in an area slightly wider than the thickness of . Second semiconductor layer 2 sandwiched between channel layers 23 and 25
As will be described later, the thickness of each channel layer 23,
The thickness is such that the spread of electrons generated in each of the regions 25 and 25 overlaps each other sufficiently. Further, the thickness of the third semiconductor layer 26 on the second channel layer 25 is such that this region in which the electrons spread does not reach the fourth semiconductor layer 27.

In other words, the energy band near the channel of the present FET formed by the channel layers 23 and 25 has the structure shown in FIG. 3(a). Further, FIG. 2B shows a semiconductor region corresponding to this energy band. That is, the energy band diagram shows, in order from the substrate surface side located on the left side of the figure, the third semiconductor layer 26, the second channel layer 25, the second semiconductor layer 24, and the first channel layer 23.
and are drawn corresponding to the first semiconductor layer 22. The central part of the band therefore corresponds to the second semiconductor layer 24 sandwiched between each channel doping plane. Here, both sides of each channel layer 23, 25, which is thinned with high concentration, are sandwiched between semiconductor layers 22, 24, 26, and each channel layer 23, 25 is sandwiched between semiconductor layers 22, 24, 26.
, 25 are formed as thin as 100 angstroms. Therefore, the conduction band is bent, and a V-shaped potential is formed corresponding to each channel layer 23, 25 as shown in the figure, thereby forming two quantum wells. Note that the thickness of each channel layer 23 and 25 is 100 angstroms, but in order to form such a quantum well in the conduction band, they must be thin to a certain extent, for example, each of 200 angstroms or less. good.

Electrons in the channel are in the lowest subband EA in the ground state with a small applied electric field. However, by applying a larger electric field to gain energy, the electrons move to the subband EB located above this subband EA, and then to subbands with higher energy levels. Here, the existence probability of electrons in each subband exhibits the spread of the wave function shown in the figure, and each channel layer 23,
In each of the semiconductor layers 22 and 26 sandwiching the channel layer 25, the value approaches zero at a portion slightly wider than the semiconductor region where the channel layers 23 and 25 are formed. Furthermore, the thickness of the second semiconductor layer 24 sandwiched between the channel layers 23 and 25 is such that the existence probability of electrons generated in each quantum well is highest at approximately the center of each doped surface in the ground state. The thickness is set to In other words, the thickness is set such that the wave function waveforms drawn corresponding to each quantum well in the ground state overlap each other sufficiently at approximately the center of each doped surface, as shown in the figure. In addition, an undoped GaAs layer on the surface side of the substrate,
In other words, the thickness of the third semiconductor layer 26 is such that electrons distributed from the second channel layer 25 to the left side in the drawing do not reach the fourth semiconductor layer 27.

Further, a gate electrode, which will be described later, is formed in Schottky contact with this fourth semiconductor layer 27, but its thickness is such that no current flows out from this gate electrode due to the tunnel effect. There is. These fourth semiconductor layer 27, third semiconductor layer 26, and second channel layer 25
, the second semiconductor layer 24 and the first channel layer 23, while satisfying the above-mentioned conditions regarding layer thickness, the total thickness thereof is sufficiently thin to satisfy the operational performance of the FET. It becomes. In addition, the uppermost contact layer 28 serves to protect the substrate surface and to serve as a drain electrode and
This is for making ohmic contact with the source electrode, and has nothing to do with the essence of the present invention.

Next, as described above, after each layer is sequentially formed on the semiconductor substrate 21, AuGe/Ni metal is formed on the uppermost contact layer 28. Then, an ohmic electrode pattern is formed using a normal photolithography technique, and a drain electrode 29 and a source electrode 30 are formed in ohmic contact with the contact layer 28 (FIG. 2(b)).
reference). Next, a gate electrode pattern is formed using a similar conventional photolithography technique. Then, using this pattern as a mask, the contact layer 28 located in the center between the drain electrode 29 and the source electrode 30 is selectively removed by etching to form a recess 31 (see FIG.
c).

Next, a Ti/Pt film in Schottky contact is formed on the fourth semiconductor layer 27 exposed in this recess 31.
A gate electrode 32 made of /Au metal is formed. As a result, an FET having the structure shown in FIG. 1 is completed.

In such a structure, the drain electrode 2
When a low voltage is applied between 9 and the source electrode 30, a low electric field is applied to the electrons in each channel layer 23, 25, and the electrons in each channel layer 23, 25 move as shown in subband EA in FIG. distribution according to the wave function waveform. In other words, there is a high probability that electrons generated from the first and second channel layers 23 and 25 will exist in the second semiconductor layer 24, which is located between these channel layers 23 and 25 and is in a substantially undoped state. Therefore, electrons in the channel are less affected by impurity scattering and travel through the channel at high speed.

As the voltage between the drain electrode 29 and the source electrode 30 is increased, each channel layer 2
The electric field strength in 3,25 increases. Therefore, subband E
The electrons in the channel, which were distributed in A, move to the subband EB, which has a higher energy level, due to the energy provided by this increase in electric field strength. Furthermore, when the drain-source voltage is increased, the electrons in the channel sequentially move to higher subbands, and eventually move from each V-type potential to the first semiconductor layer 22 and the second semiconductor layer sandwiching each channel layer 23, 25. It jumps out to the semiconductor layer 26 of No. 3. On this occasion,
The amount of electrons that jump out is overwhelmingly larger in the third semiconductor layer 26 located on the second channel layer 25, and the electrons mainly travel in this third semiconductor layer 26. Since the semiconductor layers 22 and 26, which are substantially undoped, have excellent electron transport properties, electrons travel through the channel at high speed. Therefore, the electron saturation speed does not deteriorate.

As described above, according to this embodiment, even if highly doped channel layers 23 and 25 which are susceptible to impurity scattering are used, electrons can move at high speed over the entire range from low to high electric field strength. Run through the channel with. Therefore, this FET exhibits high frequency characteristics that are equal to or better than HEMTs, and the cutoff frequency ft and maximum oscillation frequency fmax are not degraded compared to HEMTs. Furthermore, since the electron mobility in a low electric field is improved, the value of the source parasitic resistance Rs of the FET, which is affected by the electron mobility in a low electric field, is reduced. Additionally, the noise figure of the FET is reduced over the entire operating range because the electron channel velocity is increased over the entire range of field strengths.

Furthermore, in each conventional FET, the characteristics of mutual conductance gm with respect to changes in gate voltage Vg are such that the gm value peaks at a certain gate voltage value. However, in the mutual conductance characteristics according to this embodiment, the channel traveling speed of electrons increases over the entire range of electric field strength as described above, so the peak of the gm value is maintained over a wide range of gate voltage changes. has. Therefore, according to this embodiment, the design of the FET becomes easy, and the characteristics of the resulting FET become stable, making it possible to always ensure a high gain and providing an output without distortion.

Furthermore, in the FET according to this embodiment,
The fourth semiconductor layer 27 made of AlGaAs and the second channel layer 25 are located apart from each other by a distance equal to or longer than the spread of the wave function of electrons in the channel layer 25, as described above. For this reason, conventional HEMT and D
Unlike MT, deterioration of high frequency characteristics due to real space transition does not occur.

[0028] Conventionally, it was thought that a quantum well for accumulating channel electrons could only be obtained by forming a channel layer in a two-dimensional planar shape, as seen in planar-doped FETs. Therefore, the amount of impurity doped into the channel layer in a planar-doped FET is at most 1×1013 as explained in the conventional technology.
/cm2. However, according to this embodiment,
Even if the channel layer has a certain thickness, it is possible to form a quantum well as described above. Therefore, each channel layer 23, 25 of the present FET can be doped with impurities at a high concentration, and the channel is formed by a large amount of electrons. For example, the amount of impurity doped into the channel of the present FET can be as low as 8×10 13 /cm 2 per channel layer. This is because even if the doping amount per atomic layer (per 5 to 6 angstroms) is estimated as low as 5×10 12 /cm 2 , the thickness of each channel layer 23 and 25 is 100 angstroms. Therefore, in this FET, a much larger amount of channel electrons can be obtained than in a planar-doped FET, and a larger drain current can be obtained. In addition, conventional HEMTs whose current driving ability is limited by the upper limit of the two-dimensional electron gas concentration
Even compared to this, a far superior current drive capability can be obtained.

Furthermore, since the gate electrode 32 forms a Schottky contact with the fourth semiconductor layer 27 made of substantially undoped AlGaAs, the Schottky barrier becomes high. Therefore, it becomes possible to operate the device under high bias conditions, and this also improves the output characteristics.

Therefore, the FET according to this embodiment is effective when applied to the basic structure of an ultra-high frequency, high output, and low noise element.

In the above embodiment, the first, second and third semiconductor layers 22, 24 and 26 sandwiching the channel layers 23 and 25 are made of undoped GaAs, but they are not necessarily limited to this material. Not done. For example, undoped InGaAs, which has a crystal structure substantially lattice-matched to each channel layer 23, 25 and has excellent electron transport properties, may be used, and the same effect as in the above embodiment can be obtained. Further, since electrons forming the channel mainly travel through the second semiconductor layer 24 and the third semiconductor layer 26, the first semiconductor layer 22 does not necessarily have to be made of the same material as these semiconductor layers 24 and 26. Semiconductor substrate 21 and first channel layer 2
Any material having a crystal structure that is approximately lattice-matched to No. 3 is sufficient.

[0032] In addition, in the above description of the embodiment, two highly concentrated thin channel layers are provided, but it is also possible to provide only one such channel layer. The structure of such a single-layer channel FET is shown, for example, in the cross-sectional view of FIG.

A semiconductor layer 42, a channel layer 43, a semiconductor layer 44, a semiconductor layer 45, and a contact layer 46 are epitaxially grown in this order on a semi-insulating GaAs semiconductor substrate 41. The semiconductor layer 42 corresponds to the first semiconductor layer 22 in the above embodiment, similarly, the channel layer 43 corresponds to the first channel layer 23, the semiconductor layer 44 corresponds to the third semiconductor layer 26, and the semiconductor layer 45 corresponds to the fourth semiconductor layer. Layer 27 and contact layer 46 correspond to contact layer 28. That is, each of these layers shown in FIG. 4 is made of the same material and has the same thickness as the corresponding layer in the above embodiment. Further, a drain electrode 47, a source electrode 48, and a gate electrode 49 are also formed corresponding to each electrode in the above embodiment.

FIG. 5 shows the energy band near the channel of such a single layer channel structure. The left side of the figure is the substrate surface side, and the center part corresponds to the channel doping surface. Highly concentrated thin channel layer 4
3 is sandwiched between semiconductor layers 42 and 44, and the thickness thereof is thin, so that the conduction band is bent, a V-shaped potential is formed, and the illustrated quantum well is formed. Electrons in this quantum well are distributed in each subband EA
, EB and EC are distributed as shown in the wave function waveform shown in the figure. Therefore, the drain-source voltage is low,
When the electric field strength applied to the channel layer 43 is low,
Electrons in the channel reside in the lowest energy level subband EA. As understood from the waveform drawn in subband EA, the probability of the existence of electrons is
The peak appears approximately at the center of the area. Therefore, when the electric field strength in the channel is low, electrons traveling in the channel are greatly affected by impurity scattering, and their speed decreases.

However, the present FET according to the above embodiment
In this case, there is a high probability that electrons in the channel in a low electric field exist in the substantially undoped second semiconductor layer 24 that exists between the first and second channel layers 23 and 25, as described above. Therefore, in the two-layer channel FET according to this embodiment, the electron velocity is sufficiently high even in a low electric field, and the electron mobility is maintained high over the entire range of electric field strength. Therefore, this FET according to the above embodiment
In this case, the high frequency characteristics and mutual conductance gm characteristics are better, and the noise performance is also better. Moreover,
Source parasitic resistance Rs is reduced.

Note that when the electric field strength applied to the channel is high, the electron mobility of the single-layer channel FET is equal to or higher than that of the HEMT. That is, when the electric field strength increases even in a single-layer channel FET, the electrons in the channel layer 43 sequentially shift to subbands EB and EC with higher energy levels, and eventually the electrons jump out of the quantum well and become almost undoped. Semiconductor layers 42, 44
This is because you will be able to drive.

In addition, in such a single layer channel FET, since there is only one channel layer 43 thinned with high concentration, the amount of electrons forming the channel is smaller than that in the two layer channel FET according to this embodiment. There aren't many. Therefore, the FET according to this embodiment has a better current drive ability, and the FET can achieve higher output.

[0038]

[Effects of the Invention] As explained above, according to the present invention, when a low electric field is applied between the drain and the source,
Electrons generated in each of the first and second channel layers containing impurities at a high concentration are transferred to a second semiconductor layer with excellent electron transport properties that does not contain any impurities or contains impurities at a low concentration located between these channel layers. The probability of it existing increases. Furthermore, when a high electric field is applied between the drain and the source, electrons traveling in the channel gain energy, and the first and third channel layers, which have excellent electron transport properties, sandwich the first and second channel layers. jumps out to each semiconductor layer. Also, the first and second
Since each channel layer has a certain thickness, it can contain impurities at a high concentration, and the channel is formed by a large amount of electrons.

Therefore, it is possible to provide an FET with excellent current drive ability without reducing the speed of electrons traveling in the channel.

[Brief explanation of the drawing]

FIG. 1 is a sectional view showing the structure of an FET according to an embodiment of the present invention.

FIG. 2 is a process cross-sectional view showing a method of manufacturing the FET according to the embodiment shown in FIG. 1;

FIG. 3 is a diagram showing the energy band structure near the channel in the FET according to the present example.

FIG. 4 is a cross-sectional view showing the structure of a single-layer channel FET considered as a modification of this embodiment.

FIG. 5 is a diagram showing an energy band structure near the channel in the single-layer channel FET shown in FIG. 4;

FIG. 6 is a cross-sectional view showing the structure of a conventional HEMT.

FIG. 7 is a cross-sectional view showing the structure of a conventional DMT.

FIG. 8 is an energy band diagram for explaining real space transition in a conventional AlGaAs/GaAs heterojunction.

[Explanation of symbols]

21... Semi-insulating GaAs semiconductor substrate 22... First semiconductor layer (p-type undoped GaAs)
23...First channel layer (Si-doped GaAs) 24...
Second semiconductor layer (n-type undoped GaAs) 25...
Second channel layer (Si-doped GaAs) 26...Third semiconductor layer (n-type undoped GaAs) 27...Fourth semiconductor layer (n-type undoped Alx Ga1-x A
s) 28...Contact layer (n+ type GaAs) 29...Drain electrode 30...Source electrode 32...Gate electrode

Claims (1)

[Claims] 1. A first semiconductor layer made of a material with excellent electron transport properties that does not contain any impurities or contains impurities at a low concentration, and an n-type impurity that has a crystal structure that is approximately lattice-matched to the first semiconductor layer. A first channel layer made of GaAs formed thinly with a high concentration of a second semiconductor layer made of a material that is substantially lattice-matched to the second semiconductor layer, and a second channel layer made of GaAs that is thinly formed and contains a high concentration of n-type impurities and has a crystal structure that is substantially lattice-matched to the second semiconductor layer; A third semiconductor layer made of a material with excellent electron transport properties that does not contain any impurities or contains impurities at a low concentration and has a crystal structure that is substantially lattice matched to the second channel layer, and a heterojunction with the third semiconductor layer. and has an Al composition ratio X of more than 0 and less than 0.3 that does not contain any impurities or contains a low concentration of impurities that make Schottky contact with the gate electrode.
and a fourth semiconductor layer made of lx Ga1-x As.