JP5217110B2 - Semiconductor device - Google Patents

Description

  The present invention relates to a semiconductor device, and more particularly to a semiconductor device having a heterojunction bipolar transistor.

  A heterojunction bipolar transistor (hereinafter also referred to as HBT), which is a kind of bipolar transistor, is a bipolar transistor using a material having a wider band gap than the base layer for the emitter layer. Even if the impurity concentration of the base layer is increased, the injection efficiency (emitter efficiency) of electrons from the emitter layer to the base layer can be kept high. Therefore, there is a basic advantage that the emitter-base capacitance and the base resistance can be reduced and punch-through of the base layer can be suppressed without causing a decrease in current gain due to deterioration of the emitter efficiency.

  An example 74a of a conventional HBT will be described with reference to FIG. 24 (see Patent Document 1 described later).

In this heterojunction bipolar transistor 74a, on a semi-insulating GaAs substrate (not shown), an n ⁺ -type GaAs subcollector layer (not shown), an n-type GaAs collector layer 53, a p-type GaAs base layer 54, The n-type AlGaAs emitter layer 55, the n ⁺ -type GaAs emitter contact layer 55b, the base electrode 65, and the emitter electrode 66 are stacked, and the side surfaces of the emitter layer 55, the emitter contact layer 55b, and the emitter electrode 66 are covered with the insulating film 70. ing.

An intrinsic base region 54b to which carbon having a concentration of 1 × 10 ²⁰ cm ⁻³ is added is formed in the base layer on the collector layer 53. Since this carbon is passivated by hydrogen, the hole concentration is 5 × 10 ¹⁹ cm ⁻³ .

Further, an external base region (base contact region) 54c to which carbon having a concentration of 1 × 10 ²⁰ cm ⁻³ is added is formed on the surface portion around the base layer. This carbon is almost 100% activated. Therefore, the hole concentration is 1 × 10 ²⁰ cm ⁻³ . The emitter layer 55 is formed by hetero-coupling on the intrinsic base region 54b, and the base electrode 65 is formed in ohmic contact on the external base region 54c.

  FIG. 25 shows another example 74b of a heterojunction bipolar transistor (see Patent Document 2 described later).

  The heterojunction bipolar transistor 74b includes a sub-collector layer 52, a GaAs collector layer 53, a p-type AlGaAs base layer 54, an n-type AlGaAs emitter layer 55, and an n-type InGaAs emitter cap stacked on a semi-insulating GaAs substrate 51. A layer 76, an emitter contact layer 55b, a collector electrode 67, a base electrode 65, and an emitter electrode 66 are formed.

  Then, an emitter electrode 66 made of a non-alloy type electrode material is formed in the first stage of the manufacturing process of the HBT 74b, and an emitter mesa and a base electrode 65 are formed in a self-aligned (self-aligned) manner with respect to the emitter electrode 66, Further, a base and a collector mesa are formed in a self-aligned manner with respect to the base electrode 65.

  The HBT 74b has an emitter ledge structure, and the lower portion of the emitter layer 55 is extended. A base electrode 65 is formed on the extended portion, and a Pt diffusion region (base contact) 73 is formed below the extended portion. In this case, the surface of the base layer 54 is covered, and recombination on the surface can be effectively prevented.

  Since these HBTs have a high current drive capability, they are excellent as power amplifier (power amplifier: hereinafter also referred to as PA) devices, and also have the advantage of easy single power supply operation. Used as PA

  One important indicator in PA is power-added efficiency (PAE). PAE is defined as the ratio of the difference between the output power Pout and the input power Pin to the DC input power Pdc: (Pout−Pin) / Pdc. PAE is an index representing the efficiency of PA. The larger the value, the more the power consumption of the power amplifier is suppressed. In mobile communication terminals, the power consumption by the transmission-side power amplifier accounts for a large proportion of the total power consumption, so it is particularly important to increase this value.

  In order to increase the PAE, it is necessary to make the knee voltage (Vk) in the collector current Ic-collector-emitter voltage Vce characteristic of the HBT as small as possible. FIG. 26 schematically shows the Ic-Vce characteristic and the knee voltage (Vk). From this figure, in order to reduce the knee voltage (Vk), the voltage at which Ic = 0 in the Ic-Vce characteristic, that is, the so-called offset voltage (Voff) and the resistance (Ron) at the rising portion of Ic must be made as small as possible. is necessary.

  To lower Voff, the difference between the forward turn-on voltage (Vbe-on) of the emitter-base pn junction and the forward turn-on voltage (Vbc-on) of the base-collector pn junction: (Vbe -on)-(Vbc-on) needs to be as small as possible. In an HBT in which a type I heterojunction such as AlGaAs / GaAs is used for an emitter-base junction, energy discontinuity occurs at the conduction band edge, which increases Vbe-on. Therefore, for the purpose of lowering Vbe-on, a composition modulation layer (graded layer) that smoothly connects the band gap is often inserted between the emitter layer and the base layer.

  For example, in an HBT in which the base layer and the collector layer are made of a semiconductor having the same composition, when a positive voltage is applied to the base layer, the forward direction from the collector layer to the base layer is caused by hole injection from the base layer to the collector layer. Since the current increases, Voff increases. A so-called double heterojunction structure using a semiconductor having a wide band gap on the collector side is effective in suppressing such hole injection (see Non-Patent Document 1 described later).

  However, in general, a conduction band edge energy discontinuity also occurs at the base-collector junction, and particularly when the electron affinity of the collector layer is smaller than that of the base layer, the conduction band edge energy discontinuity causes a current flow. It becomes a factor which inhibits and degrades Ron. In order to avoid such a problem as much as possible, it is known to insert an undoped layer or a composition modulation layer having the same composition as the base layer between the base layer and the collector layer.

  On the other hand, in practical HBTs, from the viewpoint of ensuring reliability, as shown in Patent Document 2 described later, in order to suppress surface recombination between the emitter and the base, the emitter layer is completely formed in the base electrode formation region. In general, a so-called emitter ledge structure is used in which a part of the emitter layer is left without being removed. With such a structure, the emitter-base junction is not exposed to the surface, which helps to suppress deterioration of reliability due to an increase in surface defects.

JP-A-7-2111729 (page 5, left column, line 17 to page 5, right column, line 6, line 1) JP-A-5-136159 (page 4, right column, line 17 to page 6, left column, line 9, line 1) APPLIED PHYSICS LETTERS, Vol.83, No.26, 29 December, 2003, P5545-5547

  As described above, in the double heterostructure HBT (DHBT), by inserting the composition modulation layers between the emitter and the base and between the base and the collector, the offset voltage (Voff) generally decreases and the knee voltage (Vk) Is also reduced. However, it is necessary to optimize the structure in order to minimize the values of these parameters and at the same time to facilitate the creation of the ledge structure described above. This will be described in detail below.

  When the electron affinity of the collector layer is smaller than that of the base layer and there is a conduction band discontinuity between the base and collector, the current flow from the base layer to the collector layer is inhibited as it is. In order to avoid this, for example, consider a case where a uniform composition modulation layer is inserted between the base and collector to smoothly connect the potential of the conduction band edge between the base and collector.

  In this case, there is no particular problem when a reverse bias is applied to the pn junction between the base and collector, but when the forward bias is applied, the effect of discontinuity at the conduction band edge appears, and the knee voltage (Vk) is Increase. Such a problem may not be so important when there is a conduction band edge energy discontinuity between the emitter layer and the base layer and thermal electrons are injected from the emitter layer into the base layer. This is because electrons injected from the emitter layer to the base layer do not lose energy sufficiently in the base layer, and may exceed the peak of potential generated between the base and collector.

  However, when a uniform compositional modulation layer is inserted between the emitter and base, the average energy of electrons injected into the base layer is smaller than that in the above example, and as a result, it is generated between the base and collector. The effect of the potential mountain is greater. In other words, if a composition modulation layer is inserted between the emitter and base and between the base and collector, this alone is not sufficient to reduce Vk.

  Further, when the composition modulation layer inserted between the emitter and base is used as the emitter ledge structure described above, the following problems occur. That is, from the viewpoint of base contact formation, the thinner the layer, the better. However, as the layer becomes thinner, the emitter-base turn-on voltage (Vbe-on) increases. As a result, Voff and Vk are reduced. growing. On the contrary, when this layer is made thicker, the difficulty of forming a base contact by alloying increases.

  In the above-described double heterostructure HBT, there is a problem that the speed performance of the collector layer is generally easily deteriorated.

  The present invention has been made in view of the above situation, and an object thereof is to enable a sufficient reduction in knee voltage (Vk) in a heterojunction bipolar transistor having excellent reliability or a semiconductor device including the same. However, to provide a device structure in which base contact formation is easy and the speed performance of the collector is not easily deteriorated (and MMIC (Micro-wave-monolithic-Integrated Circuit) having this device structure as a main component) It is in.

  That is, in order to achieve the above object, the present invention provides a heterojunction bipolar transistor having at least an emitter layer, a base layer, and a collector layer (further, a subcollector layer) or a semiconductor device having the same as a main component. The emitter layer and the collector layer both have high-concentration thin layers doped with impurities at a high concentration, and the impurity concentration of these high-concentration thin layers is higher than the impurity concentration of the adjacent semiconductor layer. The present invention relates to a semiconductor device.

  According to the present invention, since both the collector layer and the emitter layer have the high-concentration thin layer having a higher impurity concentration than the adjacent semiconductor layer, respectively, the base layer, the emitter layer, and the collector layer When the composition modulation layers are inserted between the offset voltage (Voff) and the knee voltage (the offset voltage (Voff) and knee voltage ( Vk) can be sufficiently reduced, and formation of the base contact can be facilitated while suppressing a decrease in reliability.

  This effect of the present invention will be described in more detail with reference to a specific structural example as follows.

  When the electron affinity of the collector layer is smaller than that of the base layer and there is a conduction band discontinuity between the base and collector, the current flow from the base layer to the collector layer is inhibited as it is. In order to avoid this, for example, it is necessary to insert a uniform composition modulation layer between the base and the collector to smoothly connect the potential of the conduction band edge. Here, an explanation will be given taking an HBT having InP and InGaAs as main constituent materials as an example.

  FIG. 3 shows an HBT in which InP is used for the emitter layer (E), InGaAs is used for the base layer (B), and InP is used for the collector layer (C). A composition modulation layer is inserted between the emitter and base and between the base and collector. The energy band diagram (Ec: energy level of conduction band, Ev: energy level of valence band) is shown. 4A and 4B show the case where a composition modulation layer is inserted between the base and the collector (type-B HBT), and FIG. 4B is an energy band diagram when a voltage is applied. However, as shown in this figure, when the base-collector is biased in the forward direction even when the composition modulation layer is inserted, the potential barrier between the base and the collector is slightly restored, which slightly increases the Ic-Vce characteristics. Affect.

  This problem can be avoided by inserting the high-concentration thin layer together with the composition modulation layer between the base and collector, but the forward turn-on voltage (Vbc-on) of the pn junction between the base and collector is reduced, There arises a problem that the offset voltage (Voff) increases.

  Band diagrams of HBTs (type-C HBTs) in which a high-concentration thin layer is inserted between the base and collector in this way are shown in FIGS. 5A and 5B, and the type-B HBTs and type- FIG. 6 shows the Ic-Vce characteristics of C HBT. As shown in FIG. 5B, in the type-C HBT, the revival of the potential barrier between the base and the collector is suppressed even when a forward bias is applied between the base and the collector. As shown, the rise resistance is somewhat lower, but Voff increases.

  When a composition modulation layer is inserted between the emitter and base of the type-B HBT in order to lower Voff (type-D HBT), as shown in the Ic-Vce characteristics of FIG. 7, the type-B HBT Voff decreases compared to, but the rise resistance increases.

  On the other hand, when the composition modulation layer is inserted between the emitter and the base, the following problem occurs when the composition modulation layer is used as the emitter ledge described above. That is, from the viewpoint of base contact formation, the thinner the layer, the better (type-F HBT). However, as the thickness decreases, the emitter-base turn-on voltage (Vbe-on) increases and Voff increases. , Vk increases (a specific example of this will be described in an embodiment described later). Conversely, as this layer is made thicker, it becomes difficult to form a base contact by alloying the base electrode material.

  However, an increase in Voff due to such a thin composition modulation layer between the emitter and the base can be prevented by inserting the high-concentration thin layer also between the emitter and the base. FIG. 13 shows an energy band diagram of an HBT (type-G HBT) in which a high-concentration thin layer is inserted between the emitter and base in this way, and its Ic-Vce characteristic is compared with that of a type-F HBT. This is shown in FIG. As shown in this figure, the type-G HBT has an Ic-Vce characteristic in which Voff is lower than that of the type-F HBT, and it can be seen that the HBT can realize the object of the present invention.

In the present invention, the sheet carrier concentration of the high-concentration thin layer included in the emitter layer is 3 × 10 ¹¹ cm ⁻² or more and the high-concentration thin layer included in the collector layer has the above effects. The sheet carrier concentration is desirably 1.5 × 10 ¹¹ cm ⁻² or more. In this case, the sheet carrier concentration of the high concentration thin layer included in the emitter layer is preferably higher than the sheet carrier concentration of the high concentration thin layer included in the collector layer.

  The emitter layer and the collector layer may include a layer having a sum of electron affinity and band gap larger than that of the base layer. Since these layers have a function of suppressing the injection of holes from the base layer to the emitter layer or the collector layer, they are referred to herein as hole block layers, and the hole block layer on the emitter layer side is referred to as the first emitter layer. The hole block layer on the collector layer side is referred to as a first collector layer. The high-concentration thin layer included in the emitter layer is preferably in contact with the first emitter layer and made of the same material as the first emitter layer. The high concentration thin layer included in the collector layer is preferably in contact with the first collector layer and made of the same material as the first collector layer.

  Further, between the high-concentration thin layer on the emitter layer side and the base layer, and between the high-concentration thin layer on the collector side and the base layer, the semiconductor composition changes uniformly, It is desirable to insert a composition modulation layer in which the electron affinity is uniformly changed by this (modulation of the semiconductor composition).

  That is, the electron affinity of these composition modulation layers approaches the electron affinity of the base layer as it approaches the base layer, approaches the electron affinity of the hole block layer as it approaches the hole block layer, and the electron affinity at the boundary of each layer It is desirable to connect smoothly. It is good to take a structure at least close to it.

  In this case, it is preferable that the thickness of the composition modulation layer sandwiched between the emitter layer and the base layer is 15 nm or less. Since this layer is also used for the purpose of not exposing the emitter-base junction surface, it cannot be made extremely thin, but it is better as it is thinner from the viewpoint of base contact formation. It has been confirmed that about 10 nm is suitable for those purposes.

  The first collector layer may have a problem of low mobility. In this case, a second collector layer having a higher mobility than the first collector layer may be inserted between the first collector layer and the subcollector layer.

  The HBT according to the present invention preferably has the following basic configuration.

  First, the emitter layer generally includes the composition modulation layer, the high-concentration thin layer, the first emitter layer, and a second emitter to which an impurity is added at a higher concentration than the first emitter layer. The layer and the second emitter layer are made of a composition or material different from each other and are made of an emitter cap layer to which an impurity is added at a high concentration, but other layers may be inserted. For example, a third emitter layer having a higher concentration than the second emitter layer may be inserted between the second emitter layer and the emitter cap layer.

  The base layer is doped with a high concentration of impurities that produce conductivity opposite to that of the emitter layer and the subcollector layer.

  The collector layer is generally composed of the composition modulation layer, the high-concentration thin layer, and the first collector layer, but other layers may be included. For example, a second collector layer made of a composition or material different from that of the first collector layer may be inserted between the first collector layer and the subcollector layer. Further, a low-concentration layer having a constant composition may be inserted between the composition modulation layer and the base layer, or these layers may be interposed between the first collector layer and the second collector layer. A layer for smoothly connecting may be inserted.

  In this HBT, a low-resistance subcollector layer doped with impurities at a high concentration is formed on a semiconductor substrate, and the collector layer, the base layer, and the emitter layer are formed thereon, and these emitters are formed. Ohmic contact electrodes are respectively formed on the layer, the base layer, and the collector layer.

  Although it is desirable to use a semi-insulating GaAs substrate as the semiconductor substrate, InP or Si can also be used.

  Further, the MMIC according to the present invention is an integrated element including the above HBT as a main component and including passive elements such as a resistance element, a capacitance element, and an inductor.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

First Embodiment FIGS. 1 and 2 are sectional views of a double heterostructure HBT according to the present embodiment.

For example, a sub-collector layer 3 made of n ⁺ -type InGaAs and a first collector layer (hole block layer) 4 made of n ⁻ -type InP on a semiconductor substrate 1 made of semi-insulating InP via a buffer layer 2. , n ⁺ -type high-concentration thin layer 5 consisting of InP, for example, n ^- composition modulation layer 6 made of the type of AlInGaAs, p ⁺ -type base layer 7 made of InGaAs, for example, the n ^- -type about 10nm in thickness consisting AlInGaAs of composition modulation layer 8, n ⁺ -type composed of InP high density thin layer 9, n ^- first emitter layer consisting of type InP (hole blocking layer) 10, n ⁺ -type second emitter layer 11 made of InP, An emitter cap layer 12 made of n ⁺ -type InGaAs is sequentially stacked, and the surface is covered with an insulating film 16 made of Si ₃ N ₄ or the like except for the electrode forming portion.

  An emitter electrode 13 is formed on the emitter cap layer 12. Further, in order to form the base contact, the emitter cap layer 12, the second emitter layer 11, the first emitter layer 10, and the high-concentration thin layer 9 are partially removed to form a mesa structure. As shown in FIG. 1, the base electrode 14 is formed on the composition modulation layer 8 after mesa processing is performed on the surface of the base layer 7 so as to cover the surface of the composition modulation layer 8, and is further formed by heat treatment or the like. 14 may be alloyed with the compositional modulation layer 8 to form an ohmic contact portion 14a with the base layer 7, or, as shown in FIG. 2, exists in a region where the base electrode 14 is formed. The base electrode 14 may be formed after removing the composition modulation layer 8.

  The mesa structure is also formed in the formation of the collector electrode 15, and the collector electrode 15 is formed on the subcollector layer 3.

  For the emitter electrode 13, the base electrode 14, and the collector electrode 15, for example, a Ti / Pt / Au laminate can be used.

  The buffer layer 2 is made of, for example, InGaAs, InAlAs, InP, or a combination thereof.

Subcollector layer 3 has a thickness 300 to 500 nm, a concentration of ^{5 × 10 18 ~2 × 10 19} cm -3, it is also possible to use an n ⁺ -type InP besides n ⁺ -type InGaAs. The first collector layer 4 made of n ⁻ -type InP has a thickness of 100 to 700 nm, a concentration of 1 × 10 ¹⁷ cm ⁻³ or less (a sheet doping (carrier) concentration of 5 × 10 ¹² cm ⁻² or less), and n ⁺ The high-concentration thin layer 5 made of type InP has a thickness of about 5 nm or less and a sheet doping (carrier) concentration of about 5 × 10 ¹¹ cm ⁻² , and the n ⁻ -type composition modulation layer 6 has a thickness of 5 to 200 nm and a concentration It is 1 × 10 ¹⁷ cm ⁻³ or less (sheet doping (carrier) concentration is 1 × 10 ¹² cm ⁻² or less), and InGaAsP can also be used in addition to AlInGaAs.

The base layer 7 made of p ⁺ -type InGaAs has a thickness of 30 to 100 nm, a concentration of 1 × 10 ^{19 to} 5 × 10 ¹⁹ cm ⁻³ , an In composition of 0 to 0.6, and an n ⁻ type. The composition modulation layer 8 made of AlInGaAs has a thickness of about 10 nm (about 15 nm or less), a concentration of 1 × 10 ¹⁷ cm ⁻³ or less (a sheet doping (carrier) concentration of 2 × 10 ¹¹ cm ⁻² or less), and n ⁺ The high-concentration thin layer 9 made of type InP has a thickness of about 5 nm or less, and the sheet doping (carrier) concentration is about 1 × 10 ¹² cm ⁻² .

The first emitter layer 10 made of n ⁻ -type InP has a thickness of 5 to 100 nm and a concentration of 1 × 10 ¹⁷ to 1 × 10 ¹⁸ cm ⁻³ (sheet doping (carrier) concentration of 5 × 10 ¹⁰ to 1 × 10 ¹³ cm. ⁻² ), and the second emitter layer 11 made of n ⁺ -type InP has a thickness of about 50 nm and a concentration of 5 × 10 ¹⁸ to 2 × 10 ¹⁹ cm ⁻³ , and an emitter cap layer made of n ⁺ -type InGaAs. 12 has a thickness of about 50 to 100 nm and a concentration of about 1 × 10 ¹⁹ to 2 × 10 ¹⁹ cm ⁻³ .

  However, numerical values such as the film thickness and concentration of each of the above layers are typical, and are not limited to these values.

  The first emitter layer 10 and the first collector layer 4 have an electron affinity (energy difference from the reference zero level to the conduction band level Ec) and a band gap (difference between the conduction band level Ec and the valence band level Ev). It functions as a hole blocking layer whose sum is larger than that of the base layer 7.

  FIG. 13 is an energy band diagram in the direction perpendicular to the substrate (Y in the figure) corresponding to the HBT of FIG. 1 or FIG. 2, and shows the energy levels of the conduction band Ec and the valence band Ev.

  Next, the reason why the above structure has been devised will be described.

  As described above, the influence of the potential barrier generated at the base-collector interface of the type-A HBT is, for example, an InGaAs base layer and an InP collector layer as in the type-B HBT shown in FIG. 4B, when a positive voltage is applied to the base layer and the pn junction between the base and the collector is biased in the forward direction, this is shown in FIG. As you can see, the potential barrier is restored.

  This problem may not be a major problem with such type-B HBTs. This is because the electrons injected from the emitter layer into the base layer by the hot electrons are not sufficiently relaxed to the base-collector interface and may not be easily affected by the potential barrier. In other words, when a compositional modulation layer is inserted between the emitter layer and the base layer (type-D HBT), when the pn junction between the base and the collector is biased in the forward direction, FIGS. As can be inferred from this band diagram, it becomes more susceptible to the influence of this potential barrier. The results shown in FIG.

  That is, when the composition modulation layer is inserted between the emitter and the base, the necessity of inserting a high concentration thin layer together with the composition modulation layer between the base and the collector increases. In fact, the Ic-Vce characteristic of FIG. 9 shows that the knee voltage is greatly reduced in such a case (type-E HBT). Band diagrams of type-E HBTs are shown in FIGS. In this case, it can be seen that there is no potential barrier between the base and the collector.

  As described above, when the composition modulation layer is inserted between the emitter and the base as described above, even if the composition modulation layer is inserted between the base and the collector, it becomes more susceptible to the influence of the peak of the potential. The high concentration thin layer 5 is inserted between the base and the collector. The reason for this is that when a composition modulation layer is inserted between the emitter and base, the average energy of electrons injected from the emitter layer into the base layer becomes lower, making it more susceptible to the above potential peaks, This is probably because the thin layer 5 eliminates the peak of the potential, and even if electrons are injected from the emitter layer to the base layer with low energy, it can easily move from the base layer to the collector layer. Ron or Vbc-on at that time becomes an appropriate value, and Voff or Vk can be kept low.

  By the way, from the viewpoint of ensuring reliability, the HBT generally requires a structure for preventing the emitter-base interface from being exposed to the surface, that is, a so-called emitter ledge structure. That is, it is conceivable that the base contact is formed through the composition modulation layer without completely removing the composition modulation layer between the emitter and the base in the external base region.

  In this case, in order to facilitate contact formation, it is necessary to make the composition modulation layer between the emitter and base thin. For example, FIGS. 11A and 11B show band diagrams when the composition modulation layer is thinned down to 10 nm (type-F HBT). As shown in this figure, when a forward bias is applied between the emitter and the base, the potential barrier is restored between the emitter and the base. For this reason, Voff increases as shown in FIG.

  Thus, when the composition modulation layer between the emitter and the base becomes thin, it is desirable to insert a high-concentration thin layer between the emitter and the base in order to suppress the increase in Voff. 13A and 13B show band diagrams of HBT (type-G HBT) in which a high-concentration thin layer is inserted between the emitter and base as well as the composition modulation layer, and FIG. 14 shows type-G. The Ic-Vce characteristics of the HBT are compared with those of the type-F HBT. As shown in this figure, the type-G HBT provides the same level of Ic-Vce characteristics as the type-E HBT. This type-G HBT is the HBT according to the first embodiment.

As described above, the emitter-base composition modulation layer is preferably about 10 nm (about 15 nm or less) for ease of contact formation. The sheet doping concentration of the high-concentration thin layer between the emitter and base and between the base and collector is desirably 3 × 10 ¹¹ cm ⁻² or more and 1.5 × 10 ¹¹ cm ⁻² or more, respectively. This is a value obtained by simulation as the minimum concentration necessary to suppress significant deterioration of the offset voltage.

  In the present embodiment, the collector layer is constituted by the composition modulation layer 6, the high concentration thin layer 5, and the first collector layer 4, but the first collector layer 4 and the sublayer are arranged between the composition modulation layer 6 and the base layer 7. It is also possible to insert a new layer between the collector layer 3.

Second Embodiment FIG. 15 is a cross-sectional view of a double heterostructure HBT according to the present embodiment.

For example, on a semiconductor substrate 17 made of semi-insulating GaAs, n ⁺ -type subcollector layer 19, formed of GaAs of n ^- second collector layer 20 made of the type of GaAs, n ^- first consisting of type AlGaAs Collector layer (hole block layer) 21, high-concentration thin layer 22 made of n ⁺ -type AlGaAs, for example, composition modulation layer 23 made of n ⁻ -type AlGaAs, base layer 24 made of p ⁺ -type GaAs, eg n ⁻ -type composition modulation layer 25 having a thickness of about 10nm comprised of AlGaAs, n ⁺ -type high-concentration thin layer made of InGaP of 26, n ^- first emitter layer consisting of type InGaP (hole blocking layer) 27, n ^- type AlGaAs of the third emitter layer 29 made of the second emitter layer 28, n ⁺ -type GaAs, n ⁺ -type emitter cap layer 30 made of InGaAs are sequentially laminated is made of It is.

  An emitter electrode 13 is formed on the emitter cap layer 30, and an emitter cap layer 30, a third emitter layer 29, a second emitter layer 28, a first emitter layer 27, and a high-concentration thin layer 26 are formed to form a base contact. Is partially removed to form a mesa structure. As shown in FIG. 1, the base electrode 14 is formed on the composition modulation layer 25, and then an alloyed layer containing the constituent elements of the base electrode 14 and the composition modulation layer 25 as main components is formed by heat treatment. And the base layer 24, thereby forming an ohmic contact with the base layer 24. Alternatively, as shown in FIG. 2 (as shown in FIG. 15), the composition directly under the base electrode 14 is formed. The modulation layer 25 may be removed and an ohmic contact may be formed directly on the base layer 24.

  A mesa structure is also formed for forming the collector electrode 15, and the collector electrode 15 is formed on the subcollector layer 19.

  The emitter electrode 13, the base electrode 14, and the collector electrode 15 are formed from, for example, a laminate of Ti / Pt / Au.

The semiconductor surface not in contact with the electrode is covered with an insulating film 16 made of, for example, Si ₃ N ₄ .

The subcollector layer 19 is made of n ⁺ -type GaAs and has a thickness of about 300 to 500 nm and a concentration of about 5 × 10 ¹⁸ to 2 × 10 ¹⁹ cm ⁻³ . The second collector layer 20 made of n ⁻ -type GaAs has a thickness of about 100 to 700 nm and a concentration of about 1 × 10 ¹⁶ to 1 × 10 ¹⁷ cm ⁻³ , and the first collector layer 21 made of n ⁻ -type AlGaAs comprises The thickness is about 10 to 100 nm, the concentration is about 1 × 10 ¹⁶ to 1 × 10 ¹⁷ cm ⁻³ (the sheet doping concentration is about 1 × 10 ¹⁰ to 1 × 10 ¹² cm ⁻² ), and the n ⁺ type AlGaAs is used. The high-concentration thin layer 22 has a thickness of 5 nm or less and a sheet doping concentration of about 5 × 10 ¹¹ cm ^{−2. The} composition modulation layer 23 made of n ⁻ -type AlGaAs has a thickness of about 10 to 50 nm and a concentration of 1 × 10 ^16. It is about cm ⁻³ to 1 × 10 ¹⁷ cm ⁻³ (about 1 × 10 ^{10 to} 5 × 10 ¹¹ cm ^{−2 in} sheet doping concentration).

The base layer 24 made of p ⁺ type GaAs has a thickness of about 30 to 100 nm and a concentration of about 1 × 10 ^{19 to} 5 × 10 ¹⁹ cm ^{−2. The} composition modulation layer 25 made of n ⁻ type AlGaAs has a thickness of 10 nm. The high-concentration thin layer 26 made of n ⁺ -type InGaP has a thickness of about 5 nm or less and a sheet doping concentration of 1 × 10 ¹⁷ cm ⁻³ (sheet doping concentration is 2 × 10 ¹¹ cm ⁻² ) or less. It is about × 10 ¹² cm ^-2 .

The first emitter layer 27 made of n ⁻ -type InGaP has a thickness of about 5 to 100 nm and a concentration of 1 × 10 ¹⁷ to 1 × 10 ¹⁸ cm ⁻³ (1 × 10 ¹¹ to 1 × 10 ¹³ cm ^{−2 in} sheet doping concentration). The second emitter layer 28 made of n ⁻ -type AlGaAs has a thickness of about 50 to 200 nm and a concentration of about 1 × 10 ¹⁷ cm ⁻³ or more. The first emitter layer 27 and the third emitter layer 29 The Al composition is modulated so that the potential of is smoothly connected, and the concentration may not be uniform. The third emitter layer 29 made of n ⁺ -type GaAs has a thickness of about 50 nm and a concentration of about 5 × 10 ¹⁸ to 2 × 10 ¹⁹ cm ⁻³ , and the emitter cap layer 30 made of n ⁺ -type InGaAs has a thickness of 50 nm. The concentration is about 1 × 10 ¹⁹ to 2 × 10 ¹⁹ cm ⁻³ .

  FIGS. 16A and 16B are energy band diagrams in a direction perpendicular to the substrate corresponding to the HBT shown in FIG. 15, and show energy levels of the conduction band Ec and the valence band Ev.

  The first emitter layer 27 and the first collector layer 21 have an electron affinity (energy difference from the reference zero level to the conduction band level Ec) and a band gap (difference between the conduction band level Ec and the valence band level Ev). It functions as a hole blocking layer whose sum is larger than that of the base layer 24.

  In addition, although the first collector layer 21 has a problem of low mobility, the second collector layer 20 having higher mobility than the first collector layer 21 is provided between the first collector layer 21 and the subcollector layer 19. Since it is inserted, the mobility (operation speed) of the collector layer is sufficient. This is because the collector layer has a hole block layer 21 in which the sum of electron affinity and band gap is larger than that of the base layer 24, and a layer 20 in which the sum of electron affinity and band gap is smaller than that of the hole block layer 21. It is because it consists of.

  In addition, operations and effects similar to those described in the first embodiment described above can be obtained.

Third Embodiment FIG. 17 is a cross-sectional view of a double heterostructure HBT according to this embodiment.

In the present embodiment, the case where the electron affinity of the base layer is greater than that of GaAs while being formed on a GaAs substrate and using GaAs as a main component will be described. For example, on a semiconductor substrate 31 made of semi-insulating GaAs, the sub-collector layer 33 made of n ⁺ -type GaAs, n ^- collector layer made of the type of GaAs 34, n ⁺ -type high-concentration thin layer made of GaAs of 36, for example, a composition modulation layer 37 made of an n ⁻ type InGaAs layer, a base layer 38 made of p ⁺ type InGaAs, for example, a composition modulation layer 39 made of n ⁻ type AlInGaAs and having a thickness of about 10 nm, an n ⁺ type InGaP A high-concentration thin layer 40 made of n, a first emitter layer 41 made of n ⁻ -type InGaP, a second emitter layer 42 made of n-type GaAs, and an emitter cap layer 43 made of n ⁺ -type InGaAs are sequentially stacked. .

  An emitter electrode 13 is formed on the emitter cap layer 43, and a part of the emitter cap layer 43, the second emitter layer 42, the first emitter layer 41, and the high-concentration thin layer 40 is removed to form a base contact. A mesa structure is formed. As shown in FIG. 1, the base electrode 14 is formed on the composition modulation layer 39, and then an alloyed layer containing the constituent elements of the base electrode 14 and the composition modulation layer 39 as main components is formed by heat treatment. And the base layer 38, thereby forming an ohmic contact with the base layer 38. Alternatively, as shown in FIG. 2 (as shown in FIG. 17), the composition directly under the base electrode 14 is formed. The modulation layer 39 may be removed and an ohmic contact may be formed directly on the base layer 38.

  A mesa structure is also formed for forming the collector electrode 15, and the collector electrode 15 is formed on the sub-collector layer 33.

  The emitter electrode 13, the base electrode 14, and the collector electrode 15 are formed from, for example, a laminate of Ti / Pt / Au.

The semiconductor surface not in contact with the electrode is covered with an insulating film 16 made of, for example, Si ₃ N ₄ .

The sub-collector layer 33 is made of n ⁺ -type GaAs, has a thickness of about 300 to 500 nm, has a concentration of about 5 × 10 ¹⁸ to 2 × 10 ¹⁹ cm ⁻³ , and the collector layer 34 made of n ⁻ -type GaAs has a thickness of 300 to 500. About 700 nm, the concentration is about 1 × 10 ¹⁶ to 1 × 10 ¹⁷ cm ⁻³ , the high-concentration thin layer 36 made of n ⁺ -type GaAs is about 5 nm or less in thickness, and the sheet doping concentration is 5 × 10 ¹¹ cm ^−2. The composition modulation layer 37 made of n ⁻ -type InGaAs has a thickness of about 10 to 50 nm and a concentration of about 1 × 10 ¹⁶ to 1 × 10 ¹⁷ cm ⁻³ (sheet doping concentration of 1 × 10 ^{10 to} 5 × 10 ¹¹ cm ^-2 ).

The base layer 38 made of p ⁺ -type InGaAs has a thickness of about 30 to 100 nm, the concentration is about 1 × 10 ^{19 to} 5 × 10 ¹⁹ cm ⁻³ , and the composition modulation layer 39 made of n ⁻ -type AlInGaAs has a thickness of about 10 nm, The concentration is 1 × 10 ¹⁷ cm ⁻³ or less (sheet doping concentration is 2 × 10 ¹¹ cm ⁻² or less), the high-concentration thin layer 40 made of n ⁺ -type InGaP has a thickness of about 5 nm or less, and the sheet doping concentration is 1 × It is about 10 ¹² cm ⁻² .

The first emitter layer 41 made of n ⁻ -type InGaP has a thickness of about 5 to 50 nm and a concentration of 1 × 10 ¹⁷ to 1 × 10 ¹⁸ cm ⁻³ (sheet doping concentration of 5 × 10 ^{10 to} 5 × 10 ¹² cm ^−2. The second emitter layer 42 made of n-type GaAs has a thickness of about 50 to 200 nm and a concentration of about 1 × 10 ¹⁷ cm ⁻³ or more, but may not be uniform. An AlGaAs layer may be inserted between the first emitter layer 41 and the second emitter layer 42 so that the potential is smoothly connected to the first emitter layer 41. The emitter cap layer 43 made of n ⁺ -type InGaAs has a thickness of about 50 nm and a concentration of about 1 × 10 ¹⁹ to 2 × 10 ¹⁹ cm ⁻³ .

  Although the case where InGaAs is used for the base layer is shown in this embodiment mode, other materials having higher electron affinity than GaAs, for example, GaInNAs or GaAsSb can be used. In the present embodiment, the case where GaAs formed on the GaAs substrate is a main component has been shown. However, by inserting an appropriate buffer layer between the GaAs substrate 31 and the subcollector layer 33, A material having a lattice constant different from that of GaAs can be used as a main component.

  18A and 18B are energy band diagrams in a direction perpendicular to the substrate, corresponding to the HBT shown in FIG. 17, and show the energy levels of the conduction band Ec and the valence band Ev.

  In addition, the first emitter layer 41 has the sum of the electron affinity (energy difference from the reference zero level to the conduction band level Ec) and the band gap (difference between the conduction band level Ec and the valence band level Ev) of the base layer 38. It functions as a larger hole block layer.

  In addition, operations and effects similar to those described in the first embodiment described above can be obtained.

Fourth Embodiment FIGS. 19 to 23 show a schematic configuration of a DHBT according to a fourth embodiment of the present invention in which the material and layer configuration of the composition modulation layer shown in FIG. 1 or FIG. The present invention relates to a non-steep junction InP-based DHBT power amplifier in AB CDMA operation.

  First, an outline of this power amplifier will be described. In this example, an InGaAs / InAlAs superlattice layer is inserted into the emitter-base junction and the base-collector junction to remove the steep barrier in the conduction band. This non-steep junction results in a low knee voltage in the collector current-voltage characteristic, allowing high efficiency and low distortion operation.

The inventor has fabricated a large emitter-area device of 320 μm ² , for example, with an adjacent channel leakage power ratio (ACLR) of −41 dBc at 900 MHz WCDMA class AB operation and a supply voltage of 3.5 V, PAE 54%, output power 16. 0 dBm was obtained. The total collector current consumption was 21 mA. This makes it possible to realize a low supply voltage PA module for future mobile communication systems using InP related materials.

I. Introduction Next-generation mobile communication systems under development are required to simultaneously achieve high efficiency and linearity in power amplifier handsets, particularly at low output power levels. The usual approach for RF mobile systems is to use gallium-arsenide (GaAs) or silicon LDMOS for RF power amplifiers [[1] TBNishimura et al .; 2001 IEEE Radio Frequency Integrated Circuits Symposium MON2B-1. [2] DMHHartskeerl et al .; 2001 IEEE Radio Frequency Integrated Circuits Symposium RTU3A-3]. Since the double heterojunction bipolar transistor (DHBT) based on InP has a turn-on voltage lower than that of GaAs or silicon by 65%, low power supply voltage operation can be realized. InP-based double heterojunction bipolar transistors (DHBTs) have been studied for high frequency characteristics, but high efficiency characteristics in large emitter-area devices for mobile handsets in the frequency band of 800 MHz to 2100 MHz are There are few.

The inventor has found that a low knee voltage in the collector current-voltage characteristic is possible by inserting an InGaAs / InAlAs superlattice layer into these junctions and forming an appropriate doping profile. Then, using a large emitter-area device of 320 μm ² , an adjacent channel leakage power ratio (ACLR) of −41 dBc in 900 MHz W-CDMA class AB operation and a supply voltage of 3.5 V gives a PAE of 54% and an output power of 16.0 dBm. It was. The total collector current consumption was 21 mA. These results guarantee the realization of high-efficiency PA in mobile communication systems for InP-based DHBTs.

II. HBT Structure and Experiment FIG. I. A device structure in which an InGaAs / InP DHBT is provided on an InP substrate 1 ′ is shown. Superlattice layers 8 'and 6' made of InGaAs / InAlAs with InGaAs as the lower layer and InAlAs as the upper layer are inserted to remove the conduction band barriers at the emitter-base and base-collector junctions, respectively. That is, for example, 'a, a base layer 7 made of p ⁺ -InGaAs' high density thin layer 9 made of n ⁺ -InP between, and the base layer 7 'and consists of n ⁺ -InP high concentrations thin layer 5' As an AlInGaAs composition modulation layer inserted between the two layers, an Al _x In _y Ga _(1-xy) As composition modulation layer (where 0 ≦ x ≦ 1, 0 <y ≦ 1, preferably 0 ≦ x ≦ 0. 5, 0.5 ≦ y ≦ 0.6), but a plurality of layers having different composition ratios, for example, structures 8 ′ and 6 ′ in which InGaAs having a thickness of 2 nm and InAlAs having a thickness of 2 nm are stacked on each other. It is made up of.

Here, heavily doped thin InP heavily thin layers 9 'and 5' are inserted above the upper superlattice layer and below the lower superlattice layer, thereby reducing the conduction band barrier during biasing. The The total thickness of the collector layer 4 ′ is set to 500 nm in order to obtain a high breakdown voltage of 15V or more. In the figure, 3 ′ represents an n ⁺ -InGaAs subcollector layer, 4 ′ represents an n ⁻ −InP collector layer, 10 ′ represents an n ⁻ −InP first emitter layer, 11 ′ represents an n ⁺ −InP second emitter layer, Reference numeral 12 'denotes an n ⁺ -InGaAs emitter cap layer, and a buffer layer is not shown.

In constructing such devices, non-self-aligned methods and wet chemical etching were used, and Ti / Pt / Au alloyed for ohmic electrodes (not shown here) and gold-plated air A bridge interconnect is used (not shown here). DC and RF power performance was measured by using 320 μm ² emitter-area devices (4 μm × 20 μm × 4). This power performance was measured with a load pull system under efficiency matching conditions.

III. Results and discussion DC Characteristics FIG. 20 shows common emitter IV characteristics of an InGaAs / InP DHBT having a non-steep emitter-base (EB) and base-collector (BC) junction, and a steep EB junction and a non-steep BC junction, respectively. The base bias voltage was 0.70 to 0.86 V for the steep EB junction device and 0.70 to 0.78 V for the non-steep EB junction device. Here, the current when the collector current was zero was defined as Voffset, and the voltage when the collector current was 10 kA / cm ² was defined as Vx.

  A DHBT with non-steep emitter-base and base-collector junctions as shown in FIG. 19 exhibits a Voffset of 60 mV, which is 80 mV lower than that of a steep EB junction device. Vx is 234 mV, 150 mV lower than that of the steep EB junction device. This indicates that the conduction band barriers at the emitter-base and base-collector junctions are effectively removed by the insertion of the superlattice layers 8 ', 6'.

FIG. 21 shows a Gummel Plot of an InGaAs / InP DHBT of an InP DHBT having a non-steep junction. In this measurement, when the collector voltage was fixed at 1 V, a current gain (β) 93 was obtained. Despite the insertion of an InGaAs / InAlAs superlattice layer in the emitter-base junction and a slightly thick and highly doped base layer (75 nm thick, 2 × 10 ¹⁹ cm ⁻³ C doping), the above β is It is high enough for application to power amplifiers.

B. RF Small Signal Characteristics FIG. 22 shows the current cutoff frequency fT and the maximum oscillation frequency fmax as a function of the collector current for the InP DHBT of FIG. 19 applied to a 320 μm ² emitter-peripheral device. The collector bias voltage of this DHBT was fixed at 3.5 V and 1.5 V, and the decrease in fT and fmax when the bias voltage was reduced was confirmed. At the power measurement point when the collector bias voltage of this DHBT is 3.5 V and the current is 3 mA, fT is 8.8 GHz and fmax is 15.1 GHz. When the collector bias voltage was lowered to 1.5 V, fT was 8.5 GHz and fmax was 10.7 GHz. This device achieved fT and fmax sufficient for W-CDMA applications.

C. Power Performance The power performance of this InGaAs / InP DHBT (320 μm ² ) was measured at Vce = 3.5 V under the efficiency matching condition when Ic = 3 mA. Linear power efficiency and ACLR were measured as a function of input power under 900 MHz W-CDMA (Release 99) modulation. Here, the input impedance was fixed at 47.1 + j4.8Ω and the output impedance was fixed at 141−j33Ω, and the PAE under an appropriate output power was maximized.

  From the results shown in FIG. 23, the Psat point in this measurement could not be determined, but a PAE maximum value of 66.4% at an input power of 0.79 dBm was obtained. An output power of 16.0 dBm and a PAE of 54% were obtained with a low ACLR-41 dBc at 4.4 dB (back-off level from the maximum PAE value). The measured value of this PAE was 4% higher than that of the steep EB junction HBT under the efficiency matching measurement conditions. At this input level, the collector current increased to 21 mA, but was still small. These results indicate that this InP-based DHBT competes with GaAs-based DHBT in mobile communication systems and is one of the strong candidates for achieving a highly efficient and highly linear PA.

IV. CONCLUSION As described above, when a non-steep InGaAs / InP DHBT with a superlattice layer inserted at both the emitter-base and base-collector junctions is fabricated and applied to a 320 μm ² large emitter-area device, 900 MHz In W-CDMA class AB operation, it showed 54% PAE and 16.0 dBm output power at ACLR-41 dBc and supply voltage 3.5V. The improvement in PAE was due to the low Voffset and Vx reduction in the IV characteristics of non-steep junction DHBT. The combination of low collector current and high PAE indicates that InP-based HBTs are suitable for improving the performance of PAs in mobile communication systems.

  The embodiment described above can be variously modified based on the technical idea of the present invention.

For example, the impurity concentration and thickness of each layer including the above-described high concentration thin layer may be changed. About the composition modulation layer in the first to third embodiments described above, the composition ratio is changed in the range of Al _x In _y Ga _(1-xy) As (where 0 ≦ x <1, 0 <y <1). You may let me.

  Further, the shape of the mesa structure of the stacked body of the collector layer, the base layer, and the emitter layer, the arrangement of the electrodes connected to each layer, and the like are not limited to the above-described embodiments, and various shapes and arrangements can be adopted. .

  In the above-described embodiment, an npn bipolar transistor is described. However, the present invention can also be applied to a pnp type.

  It has a high current drive capability, is excellent as a power amplifier device, is easy to operate with a single power supply, is useful for power amplifiers for mobile communication terminals, etc., and can sufficiently reduce the knee voltage (Vk). However, an HBT having a device structure in which base contact formation is easy and the speed performance of the collector does not deteriorate.

It is sectional drawing of HBT by the 1st Embodiment of this invention. It is sectional drawing of HBT of another structure same as the above. It is a band figure of type-A HBT for demonstrating HBT based on this invention. It is a band figure of type-B HBT for demonstrating HBT based on this invention. It is a band figure of type-C HBT for demonstrating HBT based on this invention. It is a current-voltage characteristic view of HBT of type-B and type-C for explaining HBT based on the present invention. It is a current-voltage characteristic view of type-B and type-D HBTs for explaining an HBT based on the present invention. It is a band figure of type-D HBT for demonstrating HBT based on this invention. It is a current-voltage characteristic view of HBT of type-D and type-E for explaining HBT based on the present invention. It is a band figure of type-E HBT for demonstrating HBT based on this invention. It is a band diagram of HBT of type-F for explaining HBT based on the present invention. It is a current-voltage characteristic view of HBT of type-E and type-F for explaining HBT based on the present invention. It is a band diagram of HBT of type-G based on this invention. FIG. 6 is a current-voltage characteristic diagram of a type-G HBT and a type-F HBT according to the present invention. It is sectional drawing of HBT by the 2nd Embodiment of this invention. It is a band diagram of HBT. It is sectional drawing of HBT by the 3rd Embodiment of this invention. It is a band diagram of HBT. It is a schematic sectional drawing of HBT by the 4th Embodiment of this invention. FIG. 6 is an Ic-Vce characteristic diagram of the HBT. FIG. 6 is an I-Vb characteristic diagram of the HBT. FIG. 4 is an f-Ic characteristic diagram of the HBT. FIG. 6 is a power characteristic diagram of the HBT. It is sectional drawing of the conventional HBT. It is sectional drawing of other conventional HBT. It is a current-voltage characteristic view of HBT.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 17, 31 ... Semiconductor substrate, 2 ... Buffer layer, 3, 19, 33 ... Subcollector layer,
4, 21 ... 1st collector layer, 5, 9, 22, 26, 36, 40 ... high concentration thin layer,
6, 8, 23, 25, 37, 39 ... composition modulation layer, 7, 24, 38 ... base layer,
10, 27, 41 ... first emitter layer, 11, 28, 42 ... second emitter layer,
12, 30, 43 ... emitter cap layer, 13 ... emitter electrode, 14 ... base electrode,
14a ... ohmic contact, 15 ... collector electrode, 16 ... insulating film,
20, 34 ... second collector layer, 29 ... third emitter layer

Claims (8)

Having at least an emitter layer, a base layer, and a collector layer;
Both the collector layer and the emitter layer each have a high concentration thin layer doped with impurities at a high concentration, and the impurity concentration of these high concentration thin layers is higher than the impurity concentration of the adjacent semiconductor layer,
Between the high-concentration thin layer and the base layer included in the emitter layer and between the high-concentration thin layer and the base layer included in the collector layer, an electron affinity is obtained by modulation of a semiconductor composition. Is provided with a compositional modulation layer in which
Of these composition modulation layers, the composition modulation layer provided between the high-concentration thin layer and the base layer included in the emitter layer has a thickness of 15 nm or less, and the surface of the base layer is externally provided. The base region is also covered, and an ohmic contact portion between the base layer and the base electrode is formed by alloying with the base electrode in the external base region.
Semiconductor device. The sheet carrier concentration of the high-concentration thin layer included in the emitter layer is 3 × 10 ¹¹ cm ⁻² or more, and the sheet carrier concentration of the high-concentration thin layer included in the collector layer is 1.5 × 10 ¹¹ cm. The semiconductor device according to claim 1, which is ⁻² or more.   The semiconductor device according to claim 2, wherein a sheet carrier concentration of the high-concentration thin layer included in the emitter layer is higher than a sheet carrier concentration of the high-concentration thin layer included in the collector layer. The semiconductor device according to claim 1, wherein the composition modulation layer is made of Al _x In _y Ga _(1-xy) As (where 0 ≦ x <1, 0 <y <1). The composition modulation layer is made of Al _x In _y Ga _(1-xy) As (where 0 ≦ x ≦ 1, 0 <y ≦ 1), and is formed as a multilayer structure having a plurality of different compositions. Item 14. The semiconductor device according to Item 1.   The semiconductor device according to claim 5, wherein the composition modulation layer is a superlattice layer in which an InGaAs layer and an InAlAs layer are stacked.   2. The semiconductor device according to claim 1, wherein each of the emitter layer and the collector layer includes a hole block layer in which a sum of an electron affinity and a band gap is larger than that of the base layer.   The semiconductor device according to claim 7, wherein the high-concentration thin layer included in the emitter layer and the collector layer is formed of a material having the same composition as the hole block layer.

Images