JP2001284363A - METHOD OF MANUFACTURING InGaP/GaAs-HBT AND EVALUATING METHOD - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an evaluating method for obtaining an optimum condition by obtaining the distributions and the variations of the minority carrier lifetime, the majority carrier concentration and the thickness of a base layer between wafers and in the wafer of a InGaP/GaAs bipolar transistor HBT, and to obtain a manufacturing method thereof. SOLUTION: According to MOCVD method, a collector layer, AlGaAs layer, p-type base layer of 300 nm to 1 μm thickness, an n-InGaP layer and an emitter layer are laminated by the epitaxial growth on a GaAs substrate, the emitter layer is removed after cooling, the lifetime (τ)or minority carriers of the base layer is measured through the n-InGaP layer or directly after removing this layer, and the n-InGaP layer is removed to measure the carrier concentration (p) of the base layer and its thickness (d). The variations in lifetime τ, thickness (d) and carrier concentration (p) of the base layer are obtained for each wafer and in the wafer plane, and such conditions as the raw material gas feed distribution and temperature distribution are changed, so as to reduce the variations, thereby searching for conditions for reducing the variations.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to an InGaP / G
The present invention relates to a method for evaluating a base layer of an HBT transistor having an aAs structure and a method for manufacturing a good HBT transistor by the evaluation. HBT (Heterojunction Bipolar Trans
An istor transistor is a transistor that is most suitable for a W-CDMA transmission circuit, which is a unified standard for upcoming mobile phones. When writing InGaP / GaAs, the back means the substrate and the front means the epi layer. Since the technical fields are transistors and mobile phones, we will explain the current situation for both.

[0002]

2. Description of the Related Art There are a field effect transistor (Field Effect Transistor: FET: a kind of a unipolar transistor) and a bipolar transistor (Bipolar Transistor). FE
T utilizes the action of one type of charge (electrons or holes). It includes a channel formed in the high resistivity region, a drain electrode provided at both ends of the channel, a source electrode, and a gate electrode provided in the middle of the channel. The current is controlled by adjusting the channel width by the gate voltage and adjusting the effective flow area. The gate is cut off from the channel, the input resistance is high, and the device is a voltage amplification type device.

[0003] Bipolar transistors use two types of charges (electrons, holes). It has collector, base and emitter electrodes. an npn transistor;
There are pnp transistors. Each area has a collector (Collector), base (Base), and emitter (Emitter).
Is provided. When a current flows between the base and the emitter, a collector current several times that of the current flows. It is a current amplification type element with low input resistance. Collector current to base current ratio I
The _c / _{I b} expressed by β called current amplification factor.

In the case of a common npn Si-bipolar transistor, since the resistivity of the Si substrate is low, an n-type substrate is used as a collector, and a p-type portion as a base is formed by impurity diffusion. The mold is made by impurity diffusion. Therefore, the shape is such that the base is buried in the collector and the emitter is buried in the base. The layers are not sequentially stacked by epitaxial growth such as npn. It can be easily made by diffusion. Since all layers are Si and have different dopants, they are npn. Both pn junctions are homozygous. A pnp transistor can also be easily manufactured on a p-type Si substrate.

[0005] Bipolar transistors are already mature in Si semiconductors and are used in large quantities. The Si substrate has a low resistivity and a large single crystal of good quality is easily obtained, and the hole mobility and the electron mobility are almost the same. Si FETs have already been used in large quantities. It is a high-quality, inexpensive, and suitable semiconductor material suitable for integration. In the case of a bipolar, the base can be thinned to obtain a high speed. Also in the case of FET, high-speed response can be realized by shortening the channel.

However, since Si has a low electron mobility, it cannot be used at a high frequency of several tens of GHz or more. Since mobile phones digitize voice and transmit it on microwave frequencies, Si transistors are not enough. It has long been known that GaAs has a higher electron mobility than Si and is therefore suitable for high frequency operation. So G
A transistor using aAs as a substrate is sought.

An apparatus utilizing GaAs as a substrate and utilizing the high speed of electrons is HEMT (High Electron Mobility).
Transistor) or MESFET (Metal-Semiconductor)
Field Effect Transistor). These take advantage of the superiority of GaAs in electron mobility. This is an FET (field effect transistor)
The drain and source electrodes are provided at both ends of the n-channel, and a gate electrode is provided in the middle. Since it is an n-channel, only electrons become a problem, and GaAs electrons can be used by taking advantage of their high speed. Therefore, GaAs transistors are mature as FETs.

[0008] GaAs bipolar transistors are not mature yet and are still in the research stage. In some cases, GaAs has a low hole mobility and it is difficult to form a bipolar structure called npn. If neither electrons nor holes have similar mobilities like Si, the advantage of the bipolar transistor is small. In addition, GaAs compared to Si semiconductor
Since s-semiconductors are expensive, there must be an advantage worth it. The above is the related art of the transistor.

Next, the standard of the mobile phone will be described. The standards for mobile phones are scattered, GSM in Europe, PDC in Japan, C in the United States.
The standard called DMA1 is used. The sound emitted from the sender is converted into a digital signal, carried on a carrier signal having a constant frequency of several tens of GHz, transmitted, and transmitted to the receiver via the base station. One frequency can handle signals of 60 to 150 channels.

[0010] From the viewpoint of the sender and the receiver, the base station is not unique to a certain mobile phone but varies.
The mobile phone and the station constantly exchange signals, even when they are not on a call. So both are aware of which stations will be connected at that time. In any of the standards, the intensity of a transmission signal from a mobile phone is constant. Therefore, the amplification factor of the transistor that amplifies the transmission signal may be constant. Since the distance between the mobile phone and the station in contact at that time changes, the received signal strength at the station changes. Therefore, the amplification rate is increased for a weak signal far away from the station, and the amplification rate is reduced for a strong signal from near.

However, these portable telephones have different standards at present depending on countries and regions. Cell phones between different standards cannot communicate. There was an Iridium program that would allow calls from anywhere in the world and anywhere in the world. In this method, a signal is transmitted to an artificial satellite using microwaves, and the signal is returned from the artificial satellite to the ground for telephone communication. However, using a large number of artificial satellites requires a huge investment and increases the equipment cost, which is not feasible, and the plan is falling. The reason is that the technology of mobile phones has advanced earlier. Mobile phones should be able to communicate around the world in place of the dead iridium program.

[0012] However, it is difficult if the standards are different. For that purpose, it is necessary to unify the specifications of mobile phones. A unified standard has already been established. Unified standard is IMT
2000 and CDMA2000 (Code Division Multiple
Access) and will be launched worldwide in 2001. If worldwide standards are unified, it will be convenient to communicate around the world. Other advantages include higher signal processing speed and the ability to send images.

[0013] The new standard includes various things different from the old standard. One important thing is to vary the strength of the transmitted signal from the mobile phone. Since the combination of a mobile phone and a base station varies and the mobile phone moves, the distance between the mobile phone and the station changes. Conventionally, the transmission signal strength is fixed, but the new standard is to vary the transmission signal strength according to the distance. If it is far away, the transmission signal strength is increased, and if it is close, the signal strength is weakened. Then, the fluctuation of the intensity of the signal received by the base station is reduced, and the communication system is resistant to noise even at a long distance. In such a case, an element for amplifying the GHz signal by changing the amplification factor in a wide range is required. Since the frequency (several tens of GHz) is high, it is impossible with a Si semiconductor transistor. Even if the amplification factor is changed, the order must be variable over a wide range such as 1 to 10 times, 100 times, and 1000 times.

Although a GaAs-based semiconductor can handle a high-frequency signal, all of the devices practically used so far have been FETs (field-effect transistors). HEMT
This is because the GaAs F that takes advantage of the high speed of electron mobility
ET. Only electrons are carriers without a pn junction. This is a very convenient high-speed device. However, they are spatially limited because the current flow is parallel (channel) to the surface. The maximum current is limited by the area of the channel. Therefore, when the amplification rate is increased, signal distortion occurs. It is not possible to satisfy the condition that no distortion is generated in a wide range of the amplification factor β. I want to eliminate distortion over a wide range of 1 to 1000 times. For that purpose, a GaAs FET is useless.

[0015] Research on GaAs-bipolar transistors has long been suspended. GaAs may be difficult to form p-type and may have low hole mobility. It can be structurally difficult. Like a Si-bipolar transistor, the base is formed by p-type diffusion on a part of the collector,
It cannot be said that an emitter is formed by n-type diffusion therein. Although a p region can be created by diffusion,
It is not possible to create an n region by type diffusion. So G
In the case of aAs, an emitter layer, a base layer, and a collector layer are formed on a GaAs substrate by epitaxial growth.
Different materials are laminated to improve β. There is a heterojunction name because the materials on both sides of the pn junction are different.

A GaAs transistor in which AlGaAs is mounted on a GaAs substrate has been studied. However, this is because Al is easily oxidized and the reproducibility is poor, the operation is unstable, and practical use has been difficult. That is, there was no GaAs-bipolar transistor that could be used at a practical level. However, the unified standard CDMA for mobile phones places extremely difficult conditions on the amplification of transmission signals, so that GaAs-transistors have been attracting attention again. Although it has only been a few years, research on GaAs transistors has begun again.

Why is that? This is because new materials have been discovered. InGaP on GaAs substrate
It is to put the three-way system. This InGaP /
Since GaAs transistors do not contain Al, they have good reproducibility of epitaxial growth and high speed. It is superior to HEMT in that the signal is not distorted even if the amplification factor is increased. Since it is bipolar, the flow of current is perpendicular to the plane, and the range of current that can flow is wide. The present inventor considers that there is no element other than InGaP / GaAs that can satisfy the requirement of CDMA without distortion over a wide amplification rate.

Now, this element will be described. This is called HBT (Heterojunction Bipolar Transistor). This is the name because a pn junction can be formed at the interface between different materials. Unlike a Si-bipolar transistor, a pn junction cannot be easily formed by impurity diffusion. A pn junction can be formed by diffusion, but the interface is not sharp. Pn not suitable for high-speed operation
It will be joined. Therefore, in the case of GaAs, layers having different conductivity are stacked by epitaxial growth to form a pn junction. Since a GaAs substrate having high insulating properties is used, a collector layer (n-type) is also formed separately from the substrate. A base layer (p-type) is provided thereon, and an emitter layer is further formed.
In practice, there are other layers, resulting in a complicated epitaxial wafer. Since the base is biased and lifted, a current flows from the collector to the base and the emitter.

Among these layer structures, the base layer is the most important. First, the thickness of the base is related to the response speed. Ga
Even though As has a high electron mobility, it should not be only fast, so a thinner base thickness is better. If electrons can pass through the base while the voltage of the signal changes, there is an amplification effect,
The injected electrons do not amplify unless they can pass through the base while the signal voltage changes. If the base is thin, it can pass at the same speed in a short time. Therefore, the response speed becomes slow in inverse proportion to the base thickness. Therefore, InGaP / GaAs
In the HBT, the base thickness is as thin as 70 nm (0.07 μm). The thickness should be uniform, but it is not easy. When the thickness is thin, the fluctuation of the film thickness becomes large, and the fluctuation of the quality becomes large accordingly. In addition, the base thickness cannot be known in the manufacturing process.

The base thickness d is an important factor that determines the response speed. Further, the carrier concentration p of the base is also important. Since the base is p-type, majority carriers are holes and minority carriers are electrons. The hole density p is related to the current amplification factor. Therefore, the carrier density is also a predetermined value, and it is desired that the carrier density be small.

Another important parameter is the lifetime τ (lifetime) of electrons at the base. The flow of the base current means the flow of electrons from the emitter to the base. The electrons disappear when they recombine with the holes of the majority carriers in the base. The average time from injection into the base to recombination and disappearance is called lifetime. A short lifetime means that there is a lot of scattering and so on, which works to reduce the amplification rate. Therefore, it is also important that the life of minority (minority) carriers at the base is long and does not vary. For example, if the base has a majority carrier (hole) density of 5 × 10 ¹⁹ cm ⁻³ , the electron lifetime is 300 ps (picoseconds; ps = 10 ^−12).
s) More preferably. Thus, in the base layer,
The thickness d, hole density p, and electron lifetime τ are important parameters that govern transistor characteristics.

References regarding the HBT of InGaP / GaAs are given below. N.Pan, J.Elliot, M. Knowles, DPVu, K.Kishimot
o, JKTwynam, H. Sato, MTFresina, and GEStillm
an, "High Reliability InGaP / GaAs HBT", IEEEELECTRON
DEVICE LETTERS, vol.19, No.4, April 1998, p115

AlGaAs / GaAs or InGaP
This is a prototype of an HBT using a / GaAs-based material, and its properties were examined. It is stated that the point is to improve the quality of the carbon (C) -doped GaAs layer as the base layer in order to improve the characteristics. InGaP / GaAs HBT
States that the long-term reliability is higher than that of the AlGaAs / GaAs system. The present invention is only interested in InGaP / GaAs. AlGaAs / GaAs is not considered. This is provided to show that carbon is used as a p-type dopant and that the base layer is known to be important.

WYHan, Y. Lu, HSLee, MWCole, S.
N.Schauer, RPMoerkirk, & KAJones, "Annealing eff
ects on heavily carbon-doped GaAs ", Appl. Phys. Let
t.61 (1), 6 July 1992, p87

It is described that hydrogen is taken into the carbon-doped GaAs layer as the base layer together with carbon (C) as a dopant, thereby lowering the carrier concentration and deteriorating the crystal quality. This also mentions that the base layer of GaAs-HBT is important, and points out that hydrogen deteriorates the quality of the base layer, and is mentioned here.

As described above, HBTs having an InGaP-based epi layer mounted on a GaAs substrate have been manufactured on a trial basis. This is partly because the technology has become more mature in various aspects than when GaAs-bipolar transistors were studied 10 to 20 years ago. The quality improvement of the GaAs substrate is also a major factor. GaAs is a typical semiconductor, but cannot be easily manufactured as high as Si. GaAs is inferior in strength to Si and has a high As vapor pressure at high temperatures, so that stoichiometric (stoichometric)
iometric) difficult to do.

The GaAs single crystal was once manufactured by the HB method (Horizo
ntal Bridgman method; horizontal bridgeman method), LEC
(Liquid Encapsulate Czochralski method). In the HB method, the GaAs melt placed in a horizontal boat is cooled from the side of the seed crystal to <11.
1> Make a single crystal. This is cut at an orientation of 54.7 degrees to obtain a (100) wafer. The good points of this method are that the growth is performed while the As pressure is balanced, and that the residual stress is small. Dislocation and defect density are low because there is no As escape. Therefore, it is used as a substrate of an optical element such as an LD or an LED, which requires few dislocations and defects. However, the shape is an irregular D-shaped cross section, and if a circular wafer is formed by circular grinding, the entire peripheral portion is wasted. Expensive GaAs
Lots of material waste. Also, since it is an irregularly shaped ingot, a large circular wafer cannot be produced.

In the LEC method, the GaAs melt is covered with B ₂ O ₃ and a high pressure of an inert gas is applied to prevent As escape. There is an advantage that a circular wafer can be formed. However, since the temperature difference at the gas-liquid interface is large, strain tends to occur, and the dislocation density is large.
Although the crystal quality is not good, it can be used for devices that can be manufactured by ion implantation or the like. Used as a substrate for GaAs-HEMT or the like. By doping the isoelectronic element In, the dislocation-free GaAs ingot can be replaced with L without changing the electrical properties.
Can be made by EC method. The present applicant has the ability to pull such a high-quality LEC single crystal, but when In is included, the lattice constant is different and the price is high, which is not always good.

The applicant has invented a method called VB method (Vertical Boat method) instead of these two old methods. A GaAs single crystal is produced by placing a GaAs raw material in a cylindrical quartz tube heated by a heater and gradually cooling from a seed crystal side. Since the shape is determined by the quartz tube, a cylindrical (100)
A GaAs ingot can be manufactured. Circular wafers can be made by cutting perpendicular to the axis.

The VB method has a low distortion and a low dislocation density like the HB method. A high quality GaAs wafer can be obtained. Since this method has been developed and put into practical use by the present applicant, a high-quality GaAs substrate can be used. The LEC and VB methods, which are superior in wafer geometry (which can be circular), are currently used in parallel.
However, that situation should change in the near future. Why is that?

The diameter of the Si wafer is gradually increasing in order to increase the throughput. 8 inches (2
0 cm) to 12 inch (30 cm) wafers. Although Si ingots 30 cm in diameter and a few meters in length are huge, tremendous advances in crystallization technology have made them possible. Since the area becomes 2.5 times, the throughput is greatly improved.

The same is true for a GaAs wafer. The diameter of GaAs wafers is gradually increasing to 2 inches and 4 inches (10 cm) due to advances in crystal growth technology. But that is not enough. To reduce the cost, 5 inch and 6 inch (15 cm) G
There is a need for aAs wafers. It is difficult to lift such a large ingot by the LEC method. The temperature distribution at the solid-liquid interface causes stress on the wafer, resulting in a higher dislocation density. High quality, large diameter GaAs ingots can be manufactured by the VB method. With the current technology, only the VB method can produce such a large-diameter GaAs ingot. By growing in the (100) direction and cutting perpendicular to the axis, a large-diameter GaAs wafer can be obtained.

When a large-diameter GaAs wafer can be manufactured by the VB method, the throughput in the wafer process is improved, so that there is a possibility that an InGaP / GaAs transistor can be manufactured at low cost. No matter how high performance, even if it is expensive, it is not suitable for mobile phones and will not spread. It must be high-performance and inexpensive. Such a possibility has been opened by increasing the area of the wafer. That is, In
The conditions of peripheral technologies for manufacturing GaP / GaAs HBTs have been established.

Next, the p-type impurity will be described. GaAs
When making optical elements such as LED, LD, PD by
Zinc (Zn), magnesium (M
g) is used. These Group 2 elements are Ga (Group 3)
Since a hole is created by replacing a site, it can be a p-type dopant. Mainly Zn becomes a p-type impurity. This can easily form the p region by thermal diffusion. Mg is also a p-type dopant, but is hardly used in practice. Therefore, it can be said that the p-type impurity in the optical element is mainly Zn.

However, GaAs which is the object of the present invention
-In the HBT, Zn or Mg is not used as a p-type dopant. why? The present invention epitaxially grows a p-type layer without using thermal diffusion. Methods such as thermal diffusion may not be suitable for accurately forming a thin base layer.
Thermal diffusion controls the diffusion thickness with temperature and time. However, it has poor controllability and is not suitable for forming a thin layer. Epi growth is good for making thin layers accurately. When Zn is used as a dopant for the p-type layer, the diffusion coefficient is too large, and Zn rapidly diffuses from the base to the adjacent emitter and collector during epi growth and during the wafer process. As a result, the carrier concentration of the base layer and other layers fluctuates ex post facto. This should not be. Therefore, Zn is not used as a p-type impurity in the HBT.

Then, what is a p-type impurity? It is carbon (C). Carbon replaces the group 5 As site to generate holes and becomes a p-type impurity. It has long been known that carbon becomes p-type in GaAs, but carbon was an undesirable impurity that was difficult to control when an optical device was manufactured by liquid phase epitaxy. This is one in which carbon is eliminated as much as possible and Zn is doped. Liquid phase epi is a classical epi method, but is still used in the manufacture of optical devices (LED, LD, PD) because it can be grown near thermal equilibrium and of good quality.

However, HBT is not liquid phase epitaxy like an optical device, but MOCVD (Metallorganic Chemical V).
apor Deposition method; Organic Metal Vapor Phase epitaxy; or OMVPE method
method). Similarly, the manufacturing method is different even if it is called epi growth. Naturally, the raw materials are also different.

In MOCVD, the raw material is supplied in a gaseous state. The raw material of As is arsine (AsH ₃ ) gas, which is supplied together with H _{2 of} the carrier gas. As a raw material of Ga, a compound of alkane and gallium such as methane (CH ₄ ) and ethane (C ₂ H ₆ ) is used. The organic compound of methane and Ga is trimethylgallium (CH ₃ ) ₃ Ga (TM
G), triethylgallium (C ₂ H ₅ ) ₃ Ga (TE
G). It is liquid at room temperature, but can be made gaseous by bubbling with hydrogen or the like.

The MOCVD apparatus has a horizontal rotatable susceptor on which several GaAs wafers are mounted.
There are over a dozen (Fig. 4). There is a heater inside the susceptor or outside the apparatus to heat the susceptor and the wafer. Source gas is introduced from above toward the susceptor. The source gas comes into contact with the wafer to cause a chemical reaction to generate GaAs, which grows as a thin film on the wafer. When including In or Al in the composition,
These organometallic compound gases are used. Since zinc is not used as a dopant, (CH ₃ ) ₃ Zn (TMZ) or the like is not required.

So what to do? When trimethylgallium (TMG) and arsine (AsH ₃ ) are supplied on a GaAs substrate, a GaAs thin film always grows, and the n-type and p-type conductivity types have a ratio of raw material TMG (group 3): AsH ₃ (group 5).
Depends on Since the As pressure is high, it is necessary to prevent the escape of As. Therefore, the ratio of the third group to the fifth group is increased such as 1: 100 or 1:10, and the mixture is fed into the furnace. The mixing ratio is changed in the range of about 1: 1 to 1: 1000. Even if the ratio is changed in this way, GaAs becomes a stoichiometric thin film.

However, the higher the ratio of the latter, the more n-type, and the lower the ratio of the latter, the p-type. If the ratio of the latter (As) is high, As vacancies cannot be formed, and a high-purity n-type corresponding to the purity of the raw material is obtained. Conversely, the ratio of organometallics is
When the ratio is low, vacancy (vacancies) is easily formed, and carbon easily enters therein. A
When the s ratio approaches 1: 1, the p-type (C) concentration becomes 10
Heavily doped to ^19-10 about ²⁰ cm ^-3 are possible. However, as pointed out in the above-mentioned prior art document, hydrogen H may replace carbon C, and a predetermined p
A device is required to obtain the mold concentration. Wafer temperature T
When the ratio of group 3 to group 5 in the raw material is changed, p-type and n-type are naturally formed, and the impurity concentration can be changed. This is called auto doping.

This is because TMG is used as a Ga raw material.
Only occurs when using (trimethyl Ga), TEG
This is not the case when (triethyl Ga) is used.

Various HBT peripheral technologies have been described. Next, the variation in quality will be described. MOCVD
The reaction is cause was epitaxially grown an elaborate blow a source gas while the wafer -W ₁ was heated by placing the to _W-m rotation of the plurality (m pieces) in a circular large susceptor in the furnace. FIG. 4 shows 12 GaAs wafers W ₁ , W ₂ ,.
_j, shows an example that the placing of the ... _{W 12.} Of course, the number of simultaneously processed wafers on the susceptor is arbitrary.

Therefore, there are two cases of variation in the base characteristics. One is variation in quality between the wafer _-W 1 to _W-m. The other is quality variation depending on the location on the same wafer. Since a large number of HBTs are manufactured on one wafer, variations within the wafer are also a problem. In other words, if expressed in coordinates, (j, _xj , _yj ) could be written. j is the wafer number, (x _j , y _j )
Is a two-dimensional coordinate at W _j . Using these coordinates, the base thickness d can be expressed as d (j, _xj , _yj ), and the base carrier density can be expressed as p (j, _xj , _yj ). The minority carrier lifetime at the base can also be expressed as τ (j, _xj , _yj ). The present invention relates to a method for evaluating such properties.

[0045]

One of the factors largely affecting the characteristics of the GaAs-based HBT is the quality of the base layer. Among the emitter layer, base layer, and collector layer, base quality has the strongest relation to response characteristics and the like. Unless the thickness d, carrier concentration p, minority carrier lifetime τ, and the like are predetermined values, the base layer cannot exhibit desired properties. However, since the base layer is an extremely thin layer having a thickness of 70 nm, it is difficult to control characteristics and to evaluate characteristics. Although evaluation is difficult, in order to find optimal conditions, it is necessary to perform characteristic measurement evaluation. Once the optimum conditions are determined, manufacturing becomes possible, but evaluation is indispensable as a stage until then. Characteristic evaluation is also important when changing specifications. The invention only concerns the evaluation of the base layer.

After the wafer process, the characteristics of the transistor can be cut into individual chips, mounted on a package, and finally the characteristics can be measured by a tester. In this state, even if the amplification factor β, its temperature characteristic, the breakdown voltage, and the like are measured, the intrinsic properties of the base layer are not known. I want to evaluate only the base layer. For evaluation, the base layer must be exposed. However, the state where the base layer is exposed is different from the state of the base layer when the transistor is completed. This is the problem.

Conventionally, a single-layer base layer was grown under the same conditions as the base layer to evaluate the quality. That is, GaAs
Only the collector layer and the base layer (70 nm) are epitaxially grown on the substrate, and the layers above it are not grown. At the halfway point, the epitaxial growth is stopped and the mixture is cooled and taken out. FIG. 2 (conventional example) shows the layer structure. From the top,

[0048] 4. p-GaAs layer (base) 70 nm p = 4
× 10 ¹⁹ cm ⁻³ 3. n-GaAs layer (collector) 500 nm n = 1
× 10 ¹⁶ cm ⁻³ 2. n-GaAs layer 200 nm n = 2
× 10 ¹⁸ cm ⁻³ 1. GaAs substrate (semi-insulating SI)

Is as follows. The layer structures 1 to 4 have the same composition film thickness as the actual HBT. To make an HBT transistor, an emitter layer and the like are also epitaxially grown thereon. However, to examine the base layer, the base layer is epitaxially grown so far, cooled and taken out. Since it is necessary to prevent As escape during the cooling process, cooling is performed while blowing an As source (arsine).

The characteristics of the exposed base layer are evaluated. The method has two disadvantages. One is that the actual base sandwiched between the collector and emitter layers is different from the exposed base. During the cooling process, As enters and exits from the base, and the actual base layer differs from the base layer for inspection. The stoichiometry of the surface changes as As enters and exits. Then, the carrier density fluctuates greatly. Therefore, the actual carrier concentration cannot be measured.

In addition to the difficulty of becoming different during the cooling process,
There are also drawbacks from the base being too thin. To operate at high speed, the base layer must be thin.
This results in a thin layer of 70 nm. However, the thickness of 70 nm is too thin to accurately measure the carrier concentration (p concentration). The p concentration is determined by Hall measurement. However, when the layer is thin, Hall measurement is difficult.
It is difficult to measure the thickness correctly. The thickness is measured by measuring the step by partially etching the base layer. When the thickness is small, the measurement fluctuation is large.

Even if the carrier concentration p and the thickness d can be measured, the lifetime τ of minority carriers (electrons) cannot be measured at all. It is better that the lifetime of the minority carrier is long, but if it is long and not constant, the characteristics will vary, which is not preferable.

Why can't the life time be known if it is too thin? This is because lifetime is measured by photoluminescence. The band gap of GaAs is about 1.4 eV, which is about 0.9 μm in terms of light wavelength. Τ is measured using a laser which generates light having a higher energy in a pulsed manner. When irradiated with light, a band gap transition of electrons occurs to form a hole-electron pair. Electrons are minority carriers. This recombines with the majority carrier holes in the base to emit light. This light is called fluorescence, but the life of the minority carrier (life time) can be known by examining the decay of the fluorescence.

Although the energy of the pulsed laser light is larger than the GaAs band gap, it is absorbed by GaAs. However, since the base layer is too thin, part of the laser light passes through the base layer and reaches the collector layer (n-type). Here, electron-hole pairs may be generated. The minority carriers are holes, which recombine at the collector and emit light (Luminesc).
ence). This emission is also called fluorescence. Since the base layer is thin, light emission from the base layer and light emission from the collector layer thereunder appear to overlap. Looking at the decay of fluorescence, it is not possible to distinguish fluorescence from the collector and the base. Therefore, the minority carrier lifetime at the base cannot be measured. As described above, the life of the minority carrier at the base is an important value because it determines the response speed. I don't want to know or control this.

Regarding the thickness d, a part of the base is masked with a resist, the remaining base is removed by etching, and the step is measured with an electron microscope.

As described above, the method of examining the growth of a 70 nm base monolayer on a GaAs substrate involves the problem that the base monolayer is different from the actual base layer, and the fact that the base is thin and has a thickness d and holes. There is a problem that the density p and the electron lifetime τ cannot be measured accurately.

Variation in base thickness d, hole density p,
It is a first object of the present invention to provide an evaluation method capable of accurately measuring the electron lifetime τ. It is a second object of the present invention to provide an evaluation method capable of making a more accurate association by producing an inspection base having characteristics similar to those of an actual base. Knowing the variation in the base characteristics, the distribution and temperature distribution of the source gas are changed to reduce the variation, and G
It is a third object of the present invention to provide a method for producing aAs-HBT.

[0058]

According to the method for evaluating a base layer of the present invention, a collector layer, an AlGaAs layer, a p-type base layer having a thickness of 300 nm to 1 μm, an n-InGaP layer, and an emitter layer are formed on a GaAs substrate by MOCVD. After stacking by epi growth and cooling, the emitter layer is removed and n
The minority carrier lifetime τ of the base layer is measured directly through the -InGaP layer or excluding the n-InGaP layer, and the carrier concentration p and the base layer thickness d of the base layer are measured by removing the n-InGaP layer. The above is the evaluation method of the epi-grown wafer. The following steps are added to the production method.

For each wafer, the minority carrier lifetime τ, thickness d, and carrier concentration p of the base layer in the plane of the wafer
Is obtained, and conditions such as a source gas supply distribution and a temperature distribution are changed so as to reduce the variation, and a condition having a small variation is searched for. Using that condition, an actual InGaP / GaAs-H with a thin base layer (eg, 70 nm)
BT is manufactured.

[0060]

BEST MODE FOR CARRYING OUT THE INVENTION The HBT evaluation method of the present invention is as follows.
Instead of making a thin base of 0nm, 300nm ~ 1μ
A thick base layer of m is grown. The base layer is covered with an n-InGaP layer and an emitter layer, instead of exposing the base layer. The electron lifetime τ at the base is measured by removing the emitter layer and passing through the n-InGaP layer or directly excluding the n-InGaP layer,
When an n-InGaP layer is provided, the n-InGaP layer is removed and the carrier concentration p and the thickness d of the base layer are measured. Thus, the optimum conditions are obtained by adjusting the distribution of the flow rate of the source gas, the temperature distribution of the wafer, the height of the susceptor, the rotation speed, and the like in the MOCVD apparatus. When that condition is required, an actual H having a base thickness of 70 nm is obtained.
A BT transistor is manufactured.

There are features such as increasing the thickness of the base, not exposing the base layer, and placing an AlGaAs layer under the base layer. The rationale for each is explained.

(1) Thick Base Layer The thickness of the base layer has a strong relationship with high frequency characteristics. Since the thinner one has a faster response speed, the base thickness of GaAs-HBT is 70
It is as thin as about nm (0.07 μm). However, it cannot measure the minority carrier (electron) lifetime τ. The present invention creates a thick base layer such that all of the laser light is absorbed by the base layer. The lifetime of the minority carrier is determined by measuring photoluminescence (PL) by irradiating a strong pulsed laser beam to the base layer, and knowing the lifetime of electrons from the decay. When pulse laser light having an energy larger than the base (p-GaAs) band gap (approximately 900 nm) (approximately λ = 400 nm to 500 nm) is applied, electrons are excited from the valence band of the semiconductor to the conduction band. That is, electron-hole pairs are generated. Since the base is a p-type region, the holes are majority carriers and cannot be distinguished from other holes. The electrons are minority carriers and move around in the base. The movement is random because no electric field is applied.

Electrons collide with surrounding holes and recombine to emit band gap light and disappear. The decay of light emission (fluorescence) indicates the decay of electrons. Since the pulse laser is a pulse laser, the attenuation can be accurately tracked if the pulse interval is longer than the lifetime (life time), and the lifetime τ can be known from this. However, if the base is as thin as 70 nm, the laser light travels to the underlying collector layer, where electron-hole pairs are formed. They also recombine in a similar manner, producing fluorescence.
The two signals cannot be separated. Therefore, the electron lifetime in the thin base layer could not be measured.

However, in the present invention, since a thick base layer having a thickness of about 500 nm is formed, laser light is completely absorbed by the base layer and does not reach the base. Only the fluorescence in the base layer will be observed. Therefore, the minority carrier lifetime τ at the base can be accurately determined. The fact that the base layer is thick has such a significant meaning. The electron lifetime is an important factor that determines the current amplification factor β at a high frequency. Until now, it was not possible to accurately measure τ, so that the quality of HBT could not be controlled accurately. The method according to the invention makes it possible for the first time to measure the electronic lifetime.

As a result, there are various findings. The variation of τ between wafers is large, and
268 ps (picoseconds; ps = 10 ⁻¹² s), and high ones vary widely, such as 309 ps. However, such variation is macroscopic because it is a variation between wafers, and can be corrected by adjusting the source gas distribution and the temperature distribution. Epi-growth was performed under various conditions, and the electron lifetime on the base was measured to reduce the variation among all wafers, so that the lifetime τ> 300 ps was obtained at any position on any wafer. This was only possible if the electronic lifetime could be accurately measured and the base layer evaluated.

The measurement of the majority carrier concentration (p concentration) is 70
Although it can be formed even with a thickness of nm, it becomes more accurate when a thick layer is formed.
The carrier density is measured using the Hall effect. In this method, a current J is applied in the horizontal direction, a magnetic field B is applied vertically, an electromotive force V generated on the side end face is obtained, and the concentration is known from the ratio. In brief, p is obtained from the equation pe = JB / V (e is the elementary charge). If the base layer is thin, the end face is thin and J, V
Of the measurement becomes large. When the thickness is as thin as 70 nm, the error is large, but when the thickness is as thick as about 500 nm, the measurement accuracy is increased.

The same can be said for the thickness measurement d.
When a thin base thickness is measured, the fluctuation (deviation) of a variable having a small average value becomes large, so that the reliability of the measurement is low. Since the present invention measures the thickness of a thick base, the accuracy is improved. What is the significance of measuring the thickness from a base of 70 nm to 500 nm? This is doubtful, but will be described later in the section on the AlGaAs layer. It should be noted here that increasing the thickness d also increases the measurement accuracy of d.

(2) On the base layer, n-InGaP
In the conventional method, the growth is stopped up to the base layer, and the layer is cooled while exposing the layer. Arsine (A
It is needless to say that the wafer is cooled while sH ₃ ) is applied to the wafer.

However, the characteristics of the test base layer and the base layer in the actual HBT are still significantly different. Even if As pressure is applied, the cooling situation is different from being covered with another layer. Therefore, even if the carrier concentration p of the exposed and cooled base layer is measured, it is different from the actual carrier concentration p of the base layer of the HBT.

When AsH ₃ is used to prevent the escape of As during cooling, H generated when AsH ₃ is decomposed is taken into GaAs, and the hole carrier concentration is reduced or the quality of the crystal is changed. For this reason, even if the carrier concentration p of the exposed base layer is measured, it does not match the actual carrier concentration of the HBT base layer. The conditions are not exactly the same. In the present invention, since the base layer is covered with the same layer structure as the HBT, and is cooled in that state, there is no problem that the carrier concentration p of the base layer is inconsistent. It should be possible to know the carrier concentration of the base layer of the HBT accurately.

Covering the base layer is also useful in improving the measurement accuracy of the electron (minority carrier) lifetime τ. When cooled with the base layer exposed, As vacancies,
Although the state of Ga vacancies changes, the cross-sectional area σ of scattering of electrons n also changes in addition to the concentration p. If the scattering cross section σ is different, the lifetime τ is different. Therefore, the lifetime of the exposed base layer is different from the lifetime of the base layer covered by the upper epi layer. Since the present invention covers the base layer, the lifetime τ measured on the test wafer is equal to the actual electron lifetime τ of the HBT base layer. Cooling over the base layer in this way means that the hole concentration p and the lifetime τ
This is extremely useful in terms of accurate measurement of

(3) Providing an AlGaAs layer as a base This is not an essential condition for the present invention. This is useful for increasing the accuracy when measuring the base layer thickness d. Base layer (p
-GaAs), an AlGaAs underlayer is sandwiched immediately below, and the etching rate of GaAs / AlGaAs is close to 0 by etching with an etching solution.
Only the As layer can be removed. There is selectivity in etching with a liquid depending on the material, and there is an etching rate for GaAs (referred to as an etching rate) and an etching rate for AlGaAs different even at the same temperature concentration. P-GaAs of base layer
When a part of s is covered with the resist and is selectively etched, the GaAs in the part not covered with the resist is entirely removed, and the underlying AlGaAs remains. Removing the resist results in a step equal to the p-GaAs thickness. When this step d is measured with a microscope, the thickness d of the p-GaAs (base layer) can be determined. The underlayer is not limited to AlGaAs, but may be non-InGaP. The base is non-
In GaAs, there is a selective etchant having a different rate from that of p-GaAs. The choice of the underlayer provides a condition for accurately knowing the thickness of the base layer.

There is one problem here. The actual base thickness of the HBT is 70 nm. The base thickness measured here is 5
00 nm. It does not measure the actual HBT base layer thickness. What is the significance of measuring the thickness of the base layer made for measurement? It is necessary to state that. As mentioned above, the base thickness of the actual HBT is 70 nm, which is too thin, and the measurement error is large. When the thickness is set to about 500 nm, the thickness can be accurately measured by selective etching with the underlayer and observation with a microscope.

The purpose of the thickness measurement is not on its own, but on the macroscopic MOCV such as the distribution of the source gas and the temperature distribution.
The condition of D was to be uniform. Under the same growth conditions, the layer thickness distribution d of the base layer having an average of 500 nm
(J, _xj , _yj ; 500 nm) is a layer thickness distribution d (j, _xj , _yj ; 70 nm) in the base layer having an average of 70 nm.
Should be roughly proportional to j is the wafer number, (x _j ,
y _j ) is the in-plane coordinates on the j-th wafer. d (j,
x _{j, y} j; 500nm) is j-th wafer when generating a base layer having an average 500nm because - the _(x j, y
_j ) means the thickness. The portion that becomes thinner than the average when the base layer having an average of 500 nm is stacked is 70 n on average.
When loading m base layers, it will be M thinner than the average.
That is, there is a constant K,

[0075]

(Equation 1)

It is presumed that such a relationship will hold. If so, if the source gas amount distribution and the temperature distribution are determined and adjusted so as to minimize the fluctuation of the base thickness on the average of 500 nm, the fluctuation of the base layer thickness when the 70 nm-thick base layer is manufactured by the MOCVD method. Should be minimized. Measuring the thickness on a 500 nm basis is useful for finding the optimum conditions for MOCVD. It's a bit confusing, but that's it.

[0077]

EXAMPLE In order to evaluate the quality of a base layer in improving the characteristics of InGaP / GaAs-HBT, a sample having a layer structure as shown in FIG. 1 was prepared.

This sample was used in order from the top. 5. n-InGaAs 50 nm 4. n-GaAs 100 nm n-InGaP 50 nm 4 ′. p-GaAs 500 nm p = 4 × 10 ¹⁹ cm ⁻³ 3 ′. non-AlGaAs 100 nm 2 '. non-GaAs 100 nm GaAs substrate

It is as follows. Structure and thickness other than base layer
Is optional. The design of the HBT is, for example, as shown in FIG.
If that is the case, make a sample as shown in Fig. 1 and check it.
You. Therefore, the superstructure is the same as the actual HBT in FIG.
You. However, the base layer is 500 μm thicker than the actual HBT.
m. Carrier concentration as high as 4 × 10¹⁹c
m ^-3Of the degree. The sample structure in the case of the conventional method of FIG.
Compared with the superstructure 5, 6, 7
This is one feature. So that the base layer is not exposed
The sample configuration is very similar to the HBT. Base during cooling
Since the layer is not exposed to the outside (As atmosphere),
Carrier concentration and lifetime do not change during cooling. this
The effects have been repeatedly described.

The collector layer (2 ') was non-doped so that the carrier concentration of the base layer could be measured. The collector layer immediately below the base layer is actually n-GaAs, but when measuring the Hall of the base carrier, if there is a carrier (n; electron) in the collector, the electromotive force caused by the carrier is mixed to measure the hole concentration p. May cause errors. To avoid this, non-GaAs is used as the collector layer (2 '). Even if the base layer has a thickness of 500 nm, such a possibility exists because Hall measurement is thin. If only holes are carriers, as described above, p
e = BJ / V (e is the elementary charge, J is the transverse current density,
Vertical magnetic flux density B, V is TanmenOkoshi power), the electron is mixed _{complicated, BJ / V = (pμ p} + nμ n) e / (μ p
−μ _n ). μ _p is the hole mobility and μ _n is the electron mobility. In this case, if the respective mobilities and n are not known, p cannot be known. These parameters must also be measured by other test methods. To avoid this, the layer immediately below is non-doped.

Al for further selective etching
The GaAs layer (3 ') is inserted immediately below the base layer. As described above, this means that only p-GaAs is etched, and etching is performed with an etchant that does not etch AlGaAs. Thereby, the thickness of the p-GaAs layer can be accurately determined.

(1) Using this sample, the surface side n-I
The nGaAs (7) and the n-GaAs layer (6) are etched off, and the optical characteristics (PL; Photolumine) of the p-GaAs layer (4 ′) as the base layer are passed through the n-InGaP (5).
scence; photoluminescence measurement), and minority carriers that greatly affect HBT characteristics
er) (TRPL; time-resolved photoluminescence measurement). This measurement is based on n-InG
aP (5) can also be removed. If the base layer is thin, light will be transmitted and the PL of the underlying layer will be seen. In the present invention, since the base layer is thicker, PL
Enables the measurement of the lifetime.

As shown in FIG. 4, a plurality of Ga
As the wafer - (12 pieces in this _{figure) W} 1, W _2, ...
W _j ,... Are epitaxially grown at the same time. So P
Even if it says L measurement, the same measurement is performed about many points of all the wafers. Measurements at individual points are not significant. Variations between wafers and variations within a wafer are problems. Electronic lifetime τ by PL
The reason why is known has already been described. Electron-hole pairs are generated in the base layer by the laser light, and the electrons recombine with the holes to generate fluorescence and disappear. The decay thereof indicates the lifetime τ. Inter-wafer and intra-wafer variations of τ are calculated by processing the individual data τ (j, _xj , _yj ).

(2) Further, n-InGaP is also etched.
Using the sample turned off, the carrier concentration of the p-GaAs layer was
It was measured by Hall measurement. Hall measurement in the plane direction
Apply a current and apply a magnetic field vertically to measure the electromotive force at the end face,
The carrier concentration p is determined from the ratio. This too
All wafers W_jP (j, x)
_j, Y_jThe purpose is to examine the variation of Individual
The data doesn't make much sense. Because the base layer is thick,
Of the current J and the electromotive force V are increased. The present invention
The base layer is the same as the base layer of the actual HBT.
Because it has rear density, it can be associated with HBT
Measurement of carrier concentration is significant.

(3) The thickness d of p-GaAs was measured by step measurement using step etching utilizing the selective etching property of p-GaAs / AlGaAs. The thickness of the actual HBT is 70 nm, and the present invention produces a sample having a thickness of 500 nm, so that the base film thickness d has no direct meaning. However, as already mentioned, the average thickness is 5
All wafers -W thickness _d measured at all measurement points _j at _{00nm (j, x j, y} j; 500nm) at all the measurement points of all the wafers -W _j of the average thickness of 70nm Is considered to be proportional to the thickness d (j, _xj , _yj ; 70 nm). When epitaxial growth is performed under the same conditions, the same tendency and fluctuation are present. If a condition for forming a 500 nm base layer with a uniform thickness is searched for, a 70 nm base layer can be formed to have a uniform thickness.

As described above, each layer (mainly p-GaAs)
As a result of optimizing the growth conditions and adapting to the actual growth conditions of the HBT structure, very good HBT characteristics were obtained.

[0087]

In order to evaluate the base layer of InGaP / GaAs-HBT, a sample having a thick base layer and an HBT upper layer mounted on the base layer was prepared, and after cooling, the upper layer was removed. The characteristics of the base layer are tested through the upper layer or directly, and the minority carrier lifetime, the majority carrier concentration, and the thickness of the base layer are measured. The characteristic distribution between the wafers placed on the susceptor of the MOCVD apparatus and the characteristic distribution in the wafer are evaluated, and conditions for keeping the distribution within a predetermined range are obtained. Thus, the epi growth conditions are optimized. Compared to the method of making an epi-grown sample up to a thin base layer and evaluating the base layer, it is possible to carry out quality evaluation in accordance with the actual practice. The present invention can also perform crystal quality evaluation that cannot be evaluated by an actual HBT. The InGaP / GaAs-HBT of the present invention has excellent high-speed performance and small signal distortion even when the amplification factor is changed in a wide range. It is most suitable as a transistor required on the oscillation side in the upcoming worldwide standardized mobile phone standard CDMA. The demand is huge.

[Brief description of the drawings]

FIG. 1 shows a test of the method of the present invention in which a base layer including a thick base layer of about 500 nm is epitaxially grown to an upper layer of an HBT structure, cooled and taken out to measure the minority carrier lifetime, the majority carrier concentration and the thickness of the base layer. Sectional drawing which shows the structure of a wafer.

FIG. 2 shows a test of a conventional method in which the base layer of the HBT structure is grown epitaxially, cooled and taken out while exposing p-GaAs, and the minority carrier lifetime, the majority carrier concentration, and the thickness of the base layer are measured. Sectional drawing which shows the structure of a wafer.

FIG. 3 is a cross-sectional view illustrating an example of a layer structure of InGaP / GaAs-HBT.

FIG. 4 is a plan view showing a state in which a plurality of GaAs wafers are arranged on a susceptor for performing epitaxial growth for HBT by MOCVD.

[Explanation of symbols]

 Reference Signs List 1 GaAs substrate 2 n-GaAs layer 2 'non-GaAs layer 3 n-GaAs layer 3' non-AlGaAs layer 4 p-GaAs layer (base layer) 4 'p-GaAs layer (test base layer) 5 n-InGaP Layer 6 n-GaAs layer 7 n-InGaAs layer

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Takashi Yamada 1-1-1, Koyokita-Kita, Itami-shi, Hyogo In-house Itami Works, Sumitomo Electric Industries, Ltd. No. 1-1 F-term in Sumitomo Electric Industries, Ltd. Itami Works 4M106 AA01 AB06 BA02 BA05 BA20 CA48 CB01 CB11 5F003 AZ09 BA06 BB00 BB01 BB08 BE04 BE08 BF03 BF06 BH16 BM02 BM03 BP23 BP32 BZ03

Claims (5)

[Claims] A collector layer formed on a GaAs substrate;
A GaAs layer, a 300 nm to 1 μm thick p-GaAs base layer, an n-InGaP layer, and an emitter layer are stacked by epi-growth, cooled, and then the emitter layer is removed.
Measuring the lifetime τ of minority carriers of the base layer directly through the -InGaP layer or excluding the n-InGaP layer, measuring the carrier concentration p and the base layer thickness d of the base layer by removing the n-InGaP layer, The lifetime τ of the p-GaAs base layer between wafers and in the plane of the wafer,
Determine the distribution of the thickness d and the carrier concentration p, and change the source gas supply amount, the source gas supply distribution, the temperature distribution, the growth time, or the wafer position distribution so as to approach the uniform distribution,
InGaP / Ga with thinner p-GaAs base layer
A method for manufacturing an InGaP / GaAs-HBT, wherein optimum manufacturing conditions for As-HBT are determined. 2. A method comprising: forming a collector layer on a GaAs substrate;
A GaAs layer, a 300 nm to 1 μm thick p-GaAs base layer, an n-InGaP layer, and an emitter layer are stacked by epi-growth, cooled, and then the emitter layer is removed.
Measuring the minority carrier lifetime τ of the base layer directly through the InGaP layer or excluding the n-InGaP layer, and measuring the carrier concentration p and the base layer thickness d of the base layer by removing the n-InGaP layer; I characterized by
Evaluation method of nGaP / GaAs-HBT. 3. An AlG layer directly under the p-GaAs base layer.
providing an underlayer of aAs or InGaP to form p-GaAs
P-GaAs by an etching solution for selectively etching
3. The InGaP / GaAs-HBT according to claim 2, wherein the thickness of the p-GaAs base layer is measured by removing a part of s and exposing the underlayer to measure a step.
Evaluation method. 4. The method according to claim 2, wherein the layers below the p-GaAs base layer are non-doped and the majority carrier concentration p of the p-GaAs base layer is measured by holes without the influence of the underlayer. Evaluation method of GaAs-HBT. 5. A p-GaAs having a thickness of 300 nm to 1 μm.
3. The InGaP / GaAs-HBT according to claim 2, wherein a pulse laser beam is applied to the base layer with the base layer exposed or covered with an InGaP layer to measure the lifetime τ of the minority carrier based on p-GaAs. Evaluation method.