Patent Application
20110108885
The present invention relates to a semiconductor device and a method of manufacturing a semiconductor device. In particular, the present invention relates to semiconductor devices such as heterojunction field effect transistors using a group 3-5 compound semiconductor containing nitrogen, such as gallium nitride, and to a method of manufacturing these semiconductor devices.
Gallium nitride-based heterojunction field effect transistors can operate at high frequency and are expected to be used as switching devices that are capable of being used at high power. For example, a device having a channel that is a two-dimensional gas (2DEG) generated at an interface between n-type AlGaN and intrinsic GaN can be implemented as an AlGaN/GaN-HEMT (High Electron Mobility Transistor). Such an AlGaN/GaN-HEMT is desirably a normally-off type whose source/drain junction has high impedance even when voltage is not applied to the gate, i.e. the HEMT is desirably able to operate in an enhancement mode. As a result, the transistor can operate using a single polarity power supply and has low power consumption.
A known method for achieving transistor operation in the enhancement mode involves, for example, using a structure having a recess (groove) that is a portion of the electron supply layer (an AlGaN layer in the case of an AlGaN/GaN-HEMT) in a gate region formed to be thinner than other regions. For example, Non-Patent Document 1 discloses a normally-off AlGaN/GaN transistor that has a gate recess formed by dry etching in the AlGaN layer thereof.
By forming the groove in a portion of the AlGaN layer, the electron density of the 2DEG region facing the groove region is decreased, which enables depletion of a portion of the 2DEG at an interface between the AlGaN layer and the GaN layer. As a result, the channel can be in a disconnected state even when a gate voltage is not being applied. Therefore, a normally-off mode can be achieved in which the source/drain junction of the transistor has high impedance. When a voltage is applied to the gate electrode so that electrons are induced in the 2DEG region facing the groove region, the channel conducts, thereby achieving operation in the enhancement mode.
In the transistor of Non-Patent Document 1, the strength of the piezoelectric field generated by the AlGaN layer is known to significantly affect the 2DEG concentration at the interface. If there is a large difference in lattice constants between the AlGaN layer and the GaN layer, the resulting piezoelectric field is large and the 2DEG concentration increases. If the lattice constant difference is within a certain range, then a thicker AlGaN layer causes a larger piezoelectric filed and an increased 2DEG concentration.
In the transistor of Non-Patent Document 1, however, the inventor of the present invention has noticed a problem that the current density of the channel current cannot be made sufficiently large. In other words, while forming a thin groove in the electron supply layer (AlGaN layer) enables implementation of the enhancement mode, it also causes an intermediate level due to crystal imperfections at the bottom surface of the groove. When electrons are charged at this intermediate level due to a voltage applied to the gate electrode, the charged electrons repulse the electrons that form the 2DEG, and so the channel resistance increases and the current density of the channel decreases. When used as a switching device, operation at relatively high voltage thresholds of +1 V to +3 V is required. Therefore, due to the decrease in the channel current density, even if the threshold is +2 V, an element resistance that is sufficiently low for actual application cannot be achieved.
The current density decrease caused by the space charge at the bottom of the groove can be somewhat mitigated by distancing the groove from the 2DEG region, i.e. by forming the groove to be shallow. However, making the groove shallow is equivalent to shifting the gate threshold value in a negative direction, and this prevents realization of the normally-off operation. In other words, there is a tradeoff between increasing the channel current density and achieving the normally-off operation, i.e. increasing the gate threshold value, and so there is a limit to how much the channel current density of a switching device performing normally-off operation can be improved.
According to a first embodiment of the present invention, provided is a semiconductor device comprising a group 3-5 compound semiconductor layer; a group 3-5 compound semiconductor channel layer; a carrier supply layer that is disposed between the semiconductor layer and the channel layer and that supplies the channel layer with carriers; and a control electrode disposed on the semiconductor layer. Another example according to a first embodiment of the present invention is a semiconductor device comprising a group 3-5 compound semiconductor channel layer containing nitrogen, an electron supply layer that supplies electrons to the channel layer, a semiconductor layer that is formed on a side of the electron supply layer opposite the side facing the channel layer and that is an intrinsic or n-type group 3-5 compound semiconductor containing nitrogen, and a control electrode that is formed to contact the semiconductor layer or formed with an intermediate layer interposed between itself and the semiconductor layer.
In the first embodiment, the carrier supply layer may include a groove, and the semiconductor layer may be formed in the groove. The semiconductor device may further comprise a passivation layer that covers the carrier supply layer and that includes an open portion matching an opening of the groove. The semiconductor layer lattice may match or pseudo-lattice match with the carrier supply layer and may have a lattice constant that is greater than a lattice constant of the carrier supply layer.
In the first embodiment, the semiconductor layer may be an intrinsic semiconductor or a semiconductor with a conduction type that is the same as a conduction type of the carriers. The semiconductor layer may include nitrogen. The semiconductor layer may be an InGaN layer, an AlGaN layer, or a GaN layer. The semiconductor layer may be an InxGa1-xN layer, where 0≦x≦0.2. An insulating layer may be formed between the control electrode and the semiconductor layer. The insulating layer may be a layer including at least one insulating compound selected from a group consisting of SiOx, SiNX, SiAlXOyNz, HfOx, HfAlxOy, HfSixOy, HfNxOy, AlOx, AlNxOy, GaOx, GaOxNy, TaOx, and TiNxOy. Here, the chemical formulas including x, y, and z represent insulating compounds, and represent compounds whose elemental composition ratios are expressed as stoichiometric ratios or compounds whose elemental composition ratios are not expressed as stoichiometric ratios due to the inclusion of defects or amorphous structures.
In the first embodiment, the carrier supply layer may lattice match or pseudo-lattice match with the channel layer. The channel layer may include nitrogen. The channel layer may be a GaN layer, an InGaN layer, or an AlGaN layer. The carrier supply layer may be an AlGaN layer, an AlInN layer, or an AlN layer. The control electrode may include at least one metal selected from a group consisting of Ni, Al, Mg, Sc, Ti, Mn, Ag, Sn, Pt, and In. The carriers may be electrons.
According to a second embodiment of the present invention, provided is a method of manufacturing a semiconductor device, comprising forming a group 3-5 compound semiconductor layer on a top surface of a carrier supply layer that supplies a group 3-5 compound semiconductor channel layer with carriers; and after forming the semiconductor layer, forming a control electrode. Another example according to a second embodiment of the present invention is a method of manufacturing a semiconductor device, comprising preparing a wafer having a group 3-5 compound semiconductor channel layer containing nitrogen and an electron supply layer that supplies electrons to the channel layer and that forms a top surface of the wafer; forming a semiconductor layer, which is an intrinsic or n-type group 3-5 compound semiconductor containing nitrogen, on a top surface of the electron supply layer; and after forming the semiconductor layer, forming a control electrode.
In the second embodiment, the method of manufacturing a semiconductor device may further comprise forming a groove in the top surface of the carrier supply layer, wherein forming the semiconductor layer includes forming the semiconductor layer in the groove of the carrier supply layer. The method of manufacturing a semiconductor device may further comprise forming a passivation layer that covers the carrier supply layer; and forming an open portion in the passivation layer in a region where the groove is foamed, wherein forming the groove in the top surface of the carrier supply layer includes forming the groove by etching the carrier supply layer that is exposed by the open portion of the passivation layer. Forming the semiconductor layer in the groove of the carrier supply layer may include selectively growing an epitaxial layer that becomes the semiconductor layer in the carrier supply layer exposed by the open portion of the passivation layer. Forming the groove may include forming a mask that covers a portion of the carrier supply layer; forming another carrier supply layer on the carrier supply layer in a region not covered by the mask; and removing the mask. The semiconductor layer may be an intrinsic semiconductor or a semiconductor with a conduction type that is the same as a conduction type of the carriers, and includes nitrogen. The channel layer may include nitrogen.
According to a first embodiment of the present invention, provided is a semiconductor device comprising a group 3-5 compound semiconductor channel layer containing nitrogen, an electron supply layer that supplies electrons to the channel layer, a semiconductor layer that is formed on a side of the electron supply layer opposite the side facing the channel layer and that is an intrinsic or n-type group 3-5 compound semiconductor containing nitrogen, and a control electrode that is formed to contact the semiconductor layer or formed with an intermediate layer interposed between itself and the semiconductor layer. Furthermore, with the above configuration, the semiconductor device of the first embodiment can achieve normally-off operation and has high channel current density. In the first embodiment, the electron supply layer may include a groove and the semiconductor may be formed in the groove. The following is a description of this embodiment.
The wafer 102 may be a substrate wafer used for epitaxial growth, such as a single-crystal sapphire, silicon carbide, silicon, or gallium nitride. A commercially available wafer for epitaxial growth can be used for the wafer 102. The wafer 102 is preferably insulated, but n-types or p-types can also be used.
The buffer layer 104 is formed on the wafer 102, and can be formed of a group 3-5 compound semiconductor containing nitrogen. For example, the buffer layer 104 may be a single layer of aluminum gallium nitride (AlGaN), aluminum nitride (AlN), or gallium nitride (GaN), or may be obtained by layering a plurality of these layers. The film thickness of the buffer layer 104 is not particularly limited, but is preferably between 300 nm and 3000 nm. The buffer layer 104 can be formed using metalorganic vapor phase epitaxy (MOVPE), halide VPE, or molecular beam epitaxy (MBE). The material for forming the buffer layer 104 can be a commercially available organic metal raw material, such as trimethylgallium or trimethylindium.
The channel layer 106 is formed on the buffer layer 104, and may be a group 3-5 compound semiconductor containing nitrogen. The channel layer 106 is preferably a GaN layer, but may instead be an InGaN layer or an AlGaN layer. The film thickness of the channel layer 106 is not particularly limited, but is preferably between 300 nm and 3000 nm. The channel layer 106 can be formed using the same methods used for forming the buffer layer 104, for example.
The electron supply layer 108 supplies electrons to the channel layer 106. The electron supply layer 108 is an example of a carrier supply layer. The electron supply layer 108 is formed on the channel layer 106, and a 2DEG is formed on the channel layer 106 side of the interface between the channel layer 106 and the electron supply layer 108. The electron supply layer 108 may be formed to directly contact the channel layer 106, or may be formed with a suitable intermediate layer therebetween. The electron supply layer 108 may lattice match or pseudo-lattice match with the channel layer 106, and may be an AlGaN layer, an AlInN layer, or an AlN layer.
The film thickness of the electron supply layer 108 is determined to be within a range that is narrower than the critical film thickness estimated based on a difference in lattice constants between the channel layer 106 and the electron supply layer 108. This critical film thickness may be the film thickness for achieving mitigation of stress caused by defects in the crystal lattice resulting from stress caused by lattice mismatch. The critical film thickness depends on the Al composition and the In composition of each layer, but can be exemplified as being between 10 nm and 60 nm. The electron supply layer 108 can be formed using the same methods used for forming the buffer layer 104, for example.
The electron supply layer 108 may have a groove 110 on a surface thereof that is opposite the surface facing the channel layer 106. Forming the groove 110 in the electron supply layer 108 enables the 2DEG in the lower portion of the groove 110 to be more easily depleted. As a result, the normally-off operation of the transistor can be more easily achieved.
The film thickness of the groove 110 can be determined according to threshold values for the transistor, film thickness, and composition of the semiconductor layer 112. For example, the film thickness of the groove 110 can be between 5 nm and 40 nm. A preferable range of this film thickness is exemplified as being between 7 nm and 20 nm, a more preferable range is between 9 nm and 15 nm, and a most preferable range is between 10 nm and 13 nm.
The groove 110 can be formed by applying to the electron supply layer 108 a mask that includes an opening in a region where the groove 110 is formed, and performing anisotropic etching, such as dry etching, on the electron supply layer 108 exposed by the open portion of the mask. Any material that has etching selectivity relative to the electron supply layer 108, examples of which include photo resists, inorganic films such as SiOx, and metals, can be added to the mask. The etching gas can be a chloride-based gas such as Cl2 or CH2Cl2 and a fluoride-based gas such as CHF3 or CF4.
Instead, the groove 110 can be formed by forming the mask in a region corresponding to the groove 110 after the electron supply layer 108 is formed, forming the electron supply layer 108 with the mask in place, and then removing the mask. This mask can be SiNx or SiOx, and in this case, the selective growth method can be used. This selective growth method may be MOVPE. By limning the electron supply layer 108 to have a suitable film thickness, the groove 110 need not always be formed.
The semiconductor layer 112 is formed on a surface of the electron supply layer 108 that is opposite the surface facing the channel layer 106, and may lattice match or pseudo-lattice match with the electron supply layer 108. The semiconductor layer 112 may have a lattice constant greater than that of the electron supply layer 108. By setting the lattice constant of the semiconductor layer 112 to be greater than that of the electron supply layer 108, the semiconductor layer 112 can apply tensile stress to the electron supply layer 108.
As described above, a large difference between the lattice constants of the electron supply layer 108 and the channel layer 106 causes a larger piezoelectric field, which results in an increase in the 2DEG concentration. Furthermore, if there is a prescribed difference between the lattice constants, greater thickness of the electron supply layer 108 causes a larger piezoelectric field and a higher 2DEG concentration. In such a state, if the semiconductor layer 112 exerts tensile stress on the electron supply layer 108, the piezoelectric field caused by the difference between the lattice constants of the electron supply layer 108 and the channel layer 106 can be cancelled out, thereby decreasing the 2DEG concentration or depleting the 2DEG. As a result, even if the electron supply layer 108 is formed to be relatively thick, the normally-off operation can be realized.
The semiconductor layer 112 may be an n-type or an insulated (intrinsic) group 3-5 compound semiconductor containing nitrogen. For example, the semiconductor layer 112 can be an InGaN layer, an AlGaN layer, or a GaN layer. When an AlGaN layer having a low lattice constant is used as the electron supply layer 108, tensile stress can be effectively achieved by using a GaN layer or an InGaN layer with a higher lattice constant.
In particular, the semiconductor layer 112 may be an InxGa1-x, layer (0≦x≦0.2). In this case, larger values for x result in a higher lattice constant for the crystal and in a larger effect for cancelling out the piezoelectric field of the electron supply layer 108, so that the threshold value of the transistor can be increased. On the other hand, increasing x also degrades the crystallinity of the InGaN layer, which results in worse characteristics for the transistor. Therefore, the value of x is preferably 0≦x≦0.15, more preferably 0≦x≦0.10, and most preferably 0≦x≦0.10.
When the electron supply layer 108 includes the groove 110, the semiconductor layer 112 may be formed in the groove 110. By forming the groove 110, the normally-off operation is more easily achieved, and by forming the semiconductor layer 112 in the groove 110, the film thickness of the electron supply layer 108 at the groove 110 can be increased. Even when the groove 110 is formed in the electron supply layer 108, a separation distance can be maintained between the channel and the bottom surface of the groove 110 located at an intermediate position, and so a transistor having a higher current density than conventional normally-off transistors can be obtained.
The film thickness of the semiconductor layer 112 may be between 2 nm and 200 nm, preferably between 5 nm and 100 nm, and more preferably between 7 nm and 30 nm. The semiconductor layer 112 can be formed using MOVPE, for example. When forming the semiconductor layer 112 in a prescribed region, e.g. when forming the semiconductor layer 112 in the groove 110, the semiconductor layer 112 can be selectively formed in the prescribed region. For example, a selective growth method can be used that involves covering a region other than the prescribed region of the electron supply layer 108 with an inhibiting film that prevents epitaxial growth, using MOVPE, and then epitaxially growing the semiconductor layer 112 in the prescribed region where an opening is formed in the inhibiting film. The inhibiting film may be removed by etching, or may remain as the passivation layer 120. The inhibiting film can be a silicon nitride film or a silicon oxide film with a film thickness approximately between 10 nm and 100 nm, for example.
The insulating layer 114 can be formed on the semiconductor layer 112. By forming the insulating layer 114, the leak current from the control electrode 116 to the channel can be decreased. The insulating layer 114 may be a layer including at least one insulating compound selected from the group consisting of SiOx, SiNX, SiAlxOyNz, HfOx, HfAlxOy, HfSixOy, HfNxOy, AlOx, AlNxOy, GaOx, GaOxNy, TaOx, and TiNxOy. The chemical formulas including x, y, and z represent insulating compounds, as described above, and represent compounds whose elemental composition ratios are expressed as stoichiometric ratios or compounds whose elemental composition ratios are not expressed as stoichiometric ratios due to the inclusion of defects or amorphous structures. The insulating layer 114 can be formed using sputtering, CVD, or the like. The film thickness of the insulating layer 114 can be determined according to the dielectric constant and dielectric voltage thereof. The film thickness of the insulating layer 114 can be between 2 run and 150 nm, preferably between 5 nm and 100 nm, more preferably between 7 nm and 50 nm, and most preferably between 9 nm and 20 nm.
The control electrode 116 may be formed to contact the semiconductor layer 112. In other words, the insulating layer 114 need not be provided. Instead, the control electrode 116 may be formed on the insulating layer 114, which serves as an intermediate layer between the control electrode 116 and the semiconductor layer 112. Instead of the insulating layer 114, the intermediate layer may be formed as an intrinsic (insulating) semiconductor layer.
The control electrode 116 can include at least one metal selected from the group containing Ni, Al, Mg, Sc, Ti, Mn, Ag, Sn, Pt, and In, and among these, Al, Mg, Sc, Ti, Mn, Ag, and In are preferable. Furthermore, Al, Ti, and Mg are more preferable. The control electrode 116 can be formed using vapor deposition.
The input/output electrodes 118 are formed on the electron supply layer 108. The input/output electrodes 118 can be formed by vapor deposition of metals such as Ti and Al, for example, are then machined to have a prescribed shape using a lift-off process, and are annealed at a temperature between 700° C. and 800° C.
The passivation layer 120 covers the electron supply layer 108 in a region other than the region where the control electrode 116 and the input/output electrodes 118 are formed. The passivation layer 120 can function as a mask for the selective growth as described below, and in this case, the passivation layer 120 includes an open portion that matches the opening of the groove 110. The passivation layer 120 can be a silicon nitride film or a silicon oxide film with a film thickness between approximately 10 nm and 100 nm, for example.
The element separation region 122 is formed with the electron supply layer 108 passing therethrough, in a manner to surround the active region of the transistor. The element separation region 122 defines a region in which current flows. The element separation region 122 can be formed by forming a separating groove by etching and then implanting an insulator such as a nitride, for example. Instead, the element separation region 122 can be formed by ion implantation in the formation region using nitrogen or hydrogen.
As shown in
As shown in
As shown in
As shown in
As shown in
As shown in
As shown in
Next, a suitable mask having an opening in a region that becomes the element separation region 122 is formed, and the element separation region 122 is formed by performing selective ion implantation in the open portion of the mask. The ions implanted in the element separation region 122 may be nitrogen or hydrogen, for example, and can be any type of ion that causes the electron supply layer 108 and the channel layer 106 to serve as insulators. The semiconductor device 100 shown in
In the semiconductor device 100 and manufacturing method thereof according to the present embodiment, since the semiconductor layer 112 is formed under the control electrode 116, the channel current density can be increased while the semiconductor device 100 operates in the normally-off mode. In particular, by forming the semiconductor layer 112 to have a lattice constant greater than that of the electron supply layer 108, the piezoelectric field generated by the electron supply layer 108 is cancelled out, thereby enabling easier normally-off operation and an increase in the channel current density. Furthermore, since the semiconductor layer 112 is formed in the groove 110, the effect of the groove 110 is multiplied, enabling even easier normally-off operation and a greater increase in the channel current density.
A sapphire was prepared as the wafer 102. An epitaxial wafer to be used as an HEMT was formed by using MOVPE to sequentially layer on the wafer 102 a GaN layer as the buffer layer 104, a GaN layer as the channel layer 106, and an AlGaN layer as the electron supply layer 108. The film thickness for these three layers was respectively 100 nm, 2000 nm, and 30 μm. The Al composition of the AlGaN electron supply layer 108 was 25%.
Sputtering was used to form an SiNx layer with a film thickness of 100 nm as the passivation layer 120 on the AlGaN electron supply layer 108. The resist film 130 was formed on the SiNx passivation layer 120, and lithography was used to form the open portion 132 in the resist film 130 at a position where the groove 110 was to be formed. The dimensions of the open portion 132 were 30 μm by 2 μm.
ICP plasma etching with CHF3 gas was used to remove the SiNx passivation layer 120 exposed by the open portion 132 of the resist film 130. In this way, the SiNx passivation layer 120 having an open portion was formed. Next, the etching gas was changed to CHCl2, and the AlGaN electron supply layer 108 was etched to a depth of 20 nm. As a result, the groove 110 was formed in the electron supply layer 108.
The resist film 130 on the top surface was removed by acetone, and the wafer 102 was then moved to an MOVPE reactor in which epitaxial growth was performed until an InxGaN film (x=0.08) with a film thickness of 20 nm was selectively grown in the groove 110. The InGaN layer was not doped. In this way, an InGaN insulating (intrinsic) semiconductor layer 112 was formed.
After removing the wafer 102 from the reactor, the resist film 134 was formed and lithography was used to form the open portions 136 in the resist film 134 to have the shape of the input/output electrodes 118. Using the same method described above, the SiNx passivation layer 120 exposed by the open portions 136 was removed. Vapor deposition was then used to form a Ti/Al/Ni/Au layered film, and a lift-off process was used to create the shape of the input/output electrodes 118. Next, the wafer 102 was annealed in a nitrogen atmosphere at 800° C. for 30 seconds. In this way, a pair of input/output electrodes 118 was formed.
The resist film 138 was formed and lithography was used to form the open portion 140 in the resist film 138 on the InGaN semiconductor layer 112. The width of the open portion 140 was set to be 1.5 μm. Vapor deposition was used to form the SiOx insulating film 142 with a film thickness of 10 nm and an Ni/Au layered metal film serving as the metal film 144, and a lift-off process was used to form Ni/Au control electrodes 116 and the insulating layer 114. Furthermore, nitrogen was ion-implanted around the periphery of these elements with the resist film as a mask to form the element separation region 122. In this way, the semiconductor device 100 shown in
In the same way as in the above embodiment, an epitaxial wafer to be used as an HEMT was formed by layering on a sapphire wafer 102 a GaN buffer layer 104, a GaN channel layer 106, and an AlGaN electron supply layer 108. In the same way as in the above embodiment, the SiNX passivation layer 120, the groove 110, and the pair of input/output electrodes 118 were formed. Without forming the semiconductor layer 112 in the groove 110, the same techniques as in the above embodiment were used to form the metal film 144 that becomes the control electrode 116 and the insulating film 142 that becomes the SiOx insulating layer 114 directly on the bottom surface of the groove 110, and the insulating layer 114 and control electrode 116 were then formed. The element separation region 122 was then formed using the same techniques as in the above embodiment.
In the above embodiment, the semiconductor device 100 including the groove 110 was described. However, the groove 110 need not be included. For example, an embodiment may be the semiconductor device 200 shown in
| Number | Date | Country | Kind |
|---|---|---|---|
| 2008 072584 | Mar 2008 | JP | national |
| Filing Document | Filing Date | Country | Kind | 371c Date |
|---|---|---|---|---|
| PCT/JP2009/001207 | 3/18/2009 | WO | 00 | 12/20/2010 |
