TECHNICAL FIELD

[0001] This invention relates to transistors formed of III-Nitride materials having a nitrogen-polar (N-polar) orientation.

BACKGROUND

[0002] Currently, typical power semiconductor transistors, including devices such as power MOSFETs and insulated gate bipolar transistors (IGBTs), are fabricated with silicon (Si) semiconductor material. More recently, silicon carbide (SiC) power devices have been considered due to their superior properties. III-Nitride or III-N semiconductor devices, such as gallium nitride (GaN) devices, are now emerging as attractive candidates to carry large currents, support high voltages and to provide very low on-resistance and fast switching times.

[0003] Typical III-N high electron mobility transistors (HEMTs) and related devices are formed on III-Nitride materials grown in a group-Ill polar orientation, such as the [ 0 0 0 1 ] (C-plane) orientation. That is, the source, gate, and drain electrodes of the HEMT are formed over the group-Ill face (e.g., [ 0 0 0 1 ] face) of the III-N material layers, which is typically on an opposite side of the III-N material layers from the substrate on which the III-N layers are formed. Furthermore, III-N HEMTs are typically normally-on devices, which means that they conduct current when zero voltage is applied to the gate relative to the source. These normally on devices are known as depletion mode (D-mode) devices. However, it is more desirable in power electronics to have normally-off devices, called enhancement mode (E-mode) devices, that do not conduct substantial current at zero gate voltage and require a sufficiently positive voltage applied to the gate relative to the source in order to turn on. The use of E-mode devices in power electronics can help reduce the potential for damage to the device or to other circuit components by preventing accidental turn on of the device in case of circuit failure.

SUMMARY

[0004] In a one aspect of the invention, an N-polar III-N transistor is described. The transistor includes a III-N buffer layer having a first lattice constant, and a first III-N barrier layer having a second lattice constant on the III-N buffer layer. The transistor further includes a III-N channel layer on the III-N buffer layer, the III-N channel layer having a gate region and a plurality of access regions on opposite sides of the gate region. A compositional difference between the first III-N barrier layer and the III-N channel layer causes a conductive channel to be induced in the plurality of access regions of the III-N channel layer. The transistor also includes a source, a gate, a drain, and a second III-N barrier layer between the gate and the III- N channel layer. The second III-N barrier layer has an N-face proximal to the gate and a group-Ill face opposite the N-face, and the second III-N barrier layer has a larger bandgap than the III-N channel layer. The second lattice constant is within 0.5% of the first lattice constant.

[0005] In a second aspect of the invention, another N-polar III-N transistor is described. The transistor includes a source, a gate, a drain, a III-N buffer layer having a first lattice constant, and a first III-N barrier layer having a second lattice constant on the III-N buffer layer. The transistor also includes a III-N channel layer between the III-N buffer layer and the gate, the III-N channel layer having an N-face proximal to the gate and a group-Ill face opposite the N-face. The bandgap of the III-N channel layer is smaller than that of the first III-N barrier layer. The thickness of the III-N channel layer is less than 10 nanometers, and the sheet resistance of the conductive channel formed in an access region of the transistor is less than 400 ohms/square.

[0006] In a third aspect of the invention, another N-polar III-N transistor is described. The transistor includes a source, a gate, a drain, a III-N buffer layer having a first lattice constant, and a first III-N barrier layer having a second lattice constant on the III-N buffer layer. The transistor also includes a III-N channel layer on the first III-N barrier layer, the III-N channel layer having an N-face proximal to the gate and a group-Ill face opposite the N-face. The III-N channel layer has a smaller bandgap than the first III-N barrier layer and includes a gate region beneath the gate. The transistor further includes a second III-N barrier layer between the gate and the III-N channel layer. The second III-N barrier layer has a larger bandgap than the III- N channel layer. The second lattice constant is within 0.5% of the first lattice constant, and the transistor is configured such that a conductive channel is induced in the gate region of the III-N channel layer during a first mode of operation where a positive voltage is applied to the gate relative to the source, but not during a second mode of operation where 0 Volts is applied to the gate relative to the source.

[0007] In a fourth aspect of the invention, an enhancement-mode N-polar III-

N transistor is described. The transistor includes a source, a gate, and a drain. The transistor also includes a first III-N barrier layer and a first III-N channel layer on the first III-N barrier layer. The first III-N channel layer has an N-face proximal to the gate and a group-Ill face opposite the N-face. The first III-N channel layer also has a smaller bandgap than the first III-N barrier layer, and the III-N channel layer includes a gate region beneath the gate and access regions on opposite sides of the gate region. The transistor further includes a second III-N barrier layer between the gate and the first III-N channel layer, the second III-N barrier layer having a larger bandgap than the first III-N channel layer. The transistor also includes a second III-N channel layer over the access regions of the first III-N channel layer, the second III-N channel layer including a first conductive channel and having a smaller bandgap than the second III- N barrier layer. The transistor is configured such that when a positive voltage is applied to the gate relative to the source, a second conductive channel is induced in the gate region of the first III-N channel layer, and the first conductive channel and the second conductive channel are electrically coupled to form a conductive path extending from the source to the drain.

[0008] The transistors described herein can include one or more of the following features. A thickness and aluminum fractional composition of a portion of the second III-N barrier over the gate region of the III-N channel layer can be selected to cause the channel to be induced in the access regions without being induced in the gate region when 0V is applied to the gate relative to the source. The second III-N barrier can be over the gate region but not over the plurality of access regions of the III-N channel layer. A thickness and aluminum fractional composition of the second III-N barrier can be selected such that in a first mode of operation where 0V is applied to the gate relative to the source, the conductive channel is induced in the plurality of access regions without being induced in the gate region. The transistor can be configured such that in a second mode of operation where a positive voltage is applied to the gate relative to the source, the conductive channel is induced in the gate region and in the plurality of access regions of the III-N channel layer. The second III-N barrier layer can have a third lattice constant within 0.5% of the first lattice constant. [0009] The first III-N barrier layer can include an Al _x Ini_ _x N layer, and x can be between 82 and 83 or between 81 and 84. The transistor can further include an Al _z Gai_ _z N or A1N layer. The Al _z Gai_ _z N or A1N layer can be between the Al _x Ini_ _x N layer and the III-N channel layer. The thickness of the III-N channel layer can be less than 10 nanometers. The transistor can be configured such that in a first mode of operation where 0V is applied to the gate relative to the source, the conductive channel is induced in the plurality of access regions without being induced in the gate region of the III-N channel layer, and in a second mode of operation where a positive voltage is applied to the gate relative to the source, the conductive channel is induced in the gate region and in the plurality of access regions of the III-N channel layer. The III-N buffer layer can comprise GaN. The III-N channel layer can comprise GaN. The first III-N barrier layer can comprise Al _x In _y Gai- _x - _y N, and the ratio x/y can be between 82/18 and 83/17 or between 81/19 and 84/16. The transistor can further comprise an Al _z Gai_ _z N or A1N layer. The Al _z Gai_ _z N or A1N layer can be between the Al _x In _y Gai_ _x _ _y N layer and the III-N channel layer. The thickness of the III-N channel layer can be less than 10 nanometers.

[0010] The second lattice constant can be within 0.5% of the first lattice constant. The transistor can further include a second III-N barrier layer between the gate and the III-N channel layer, wherein the second III-N barrier layer has a larger bandgap than the III-N channel layer. The second III-N barrier can be over a gate region of the transistor but not over the access region of the transistor. The thickness and aluminum fractional composition of the second III-N barrier can be selected such that in a first mode of operation where 0V is applied to the gate relative to the source, the conductive channel is induced in the plurality of access regions without being induced in the gate region. The transistor can be configured such that in a second mode of operation where a positive voltage is applied to the gate relative to the source, the conductive channel is induced in the gate region and in the plurality of access regions of the III-N channel layer. The first III-N barrier layer can comprise an Al _x Ini_ _x N or Al _x In _y Gai_ _x _ _y N layer. The threshold voltage of the transistor can be greater than 2 Volts. The breakdown voltage of the transistor can be greater than 300 Volts.

[0011] The III-N channel layer can be a first III-N channel layer, and the transistor can further comprise a second III-N channel layer having a smaller bandgap than the second III-N barrier layer, wherein the second III-N barrier layer is between the first III-N channel layer and the second III-N channel layer. The second III-N channel layer can include a recess or an aperture, and the gate can be in the recess or aperture. The compositional difference between the second III-N barrier layer and the second III-N channel layer can cause a second conductive channel to be induced in the second III-N channel layer. During the first mode of operation the conductive channel in the gate region of the first III-N channel layer and the second conductive channel in the second III-N channel layer can form a continuous electrically conductive path extending from the source to the drain. The transistor can further comprise a III-N cap layer, for example and AlInN cap layer, over the second III-N channel layer. The III-N buffer layer or the first III-N channel layer can comprise GaN. The second III-N channel layer can comprise GaN. The transistor can be configured such that when the transistor is in the ON state and conducting current from the drain to the source, the current flows through both the first and second conductive channels.

[0012] The details of one or more embodiments of the subject matter described in this specification are set forth in the accompanying drawings and the description below. Other features, aspects, and advantages of the subject matter will become apparent from the description, the drawings, and the claims.

DESCRIPTION OF DRAWINGS

[0013] Figures 1 and 2 are cross-sectional views of N-polar Ill-Nitride enhancement-mode transistors.

[0014] Figure 3A is a cross-sectional view of an N-polar Ill-Nitride enhancement-mode transistor while biased in the OFF state.

[0015] Figure 3B is a cross-sectional view of the transistor of Figure 3 A while biased in the ON state.

[0016] Like reference symbols in the various drawings indicate like elements.

DETAILED DESCRIPTION

[0017] Described herein are Ill-Nitride (i.e., III-N) transistors, and in particular enhancement-mode (E-mode) III-N transistors, for which the III-N material is oriented in an N-polar direction. The N-polar (i.e., N-face) III-N material structure is prepared either by directly growing N-polar III-N layers on a suitable substrate such as silicon, Silicon Carbide (SiC), sapphire, Aluminum Nitride (ΑΓΝ), or Gallium Nitride (GaN), or by growing the layers in reverse order in a group-Ill polar orientation and then accessing an N-face of the III-N material structure from the reverse side of the direction of growth, for example by removing the growth substrate and optionally by removing a portion of the III-N material that was directly adjacent to the growth substrate. As used herein, the terms Ill-Nitride or III-N materials, layers, devices, etc., refer to a material or device comprised of a compound semiconductor material according to the stoichiometric formula B _w Al _x In _y Ga _z N, where w+x+y+z is about 1, 0 < w < 1, 0 < x < 1, 0 < y < 1, and 0 < z < 1. In a Ill-Nitride or III-N device, such as a transistor or HEMT, the conductive channel can be partially or entirely contained within a III-N material layer.

[0018] An exemplary N-polar Ill-Nitride E-mode device is shown in Figure 1.

Layers 12, 14, 16, and 18 are N-polar III-N layers, oriented as shown in the [ 0 0 0 -1 ] direction. The III-N layers 12, 14, 16, and 18 are optionally on a substrate 10 such as silicon, SiC, sapphire, A1N, GaN, or other substrate material suitable for epitaxial growth of Ill-Nitride layers. In some implementations the substrate 10 may be omitted. A gate 23 is over III-N layer 18, with an insulating layer 20 optionally included between gate 23 and layer 18. As seen, the source and drain contacts 21 and 22, as well as the gate 23, are all formed over the N-face (e.g., [ 0 0 0 -1 ] face) of the III-N material layers, which is on an opposite side of the III-N material layers from the substrate 10. Source and drain contacts 21 and 22, respectively, are on opposite sides of the gate 23 and contact the device 2DEG channel 19 that is formed in layer 16. The portions of the III-N materials directly below the gate 23 are referred to as the gate region of the device. The portions of the III-N materials directly below the source and drain 21 and 22, are respectively referred to as the source and drain regions of the device. The portions of III-N material between the gate region and the source region, and between the gate region and the drain region, are referred to as the device access regions.

[0019] Layer 12 is a III-N buffer layer, which can for example be GaN. The buffer layer 12 can be rendered insulating or substantially free of n-type mobile carriers by including dislocations or point defects in the layer, or by doping the layer with compensating elements, such as Fe, C, and/or Mg. The buffer layer can have a substantially uniform composition throughout, or the composition can vary. For example, in some implementations the buffer layer is graded, such as by grading the aluminum composition in the buffer layer. The buffer layer 12 can be substantially thicker than any of the other Ill-Nitride layers in the structure. For example, buffer layer 12 may have a thickness that is at least 10 times, but typically at least 30 times, the combined thickness of the III-N layers between buffer layer 12 and the gate 23.

[0020] Layer 14 is a back-barrier III-N layer (i.e., a first III-N barrier layer).

Layer 16 is a III-N channel layer, which can be formed of GaN. Layer 18 is a gate- barrier III-N layer (i.e., a second III-N barrier layer). The back-barrier III-N layer 14 has a larger bandgap than the channel layer 16, such that a 2DEG channel 19 is induced in the access regions of the channel layer 16 adjacent the interface between the back-barrier III-N layer 14 and the channel layer 16.

[0021] The portion of the 2DEG in the access regions is populated with conductive charge both when the gate 23 is biased relative to the source 21 at a voltage greater than the device threshold voltage, and when the gate 23 is biased relative to the source 21 at a voltage less than the device threshold voltage. In order to prevent surface roughening at the surface of channel layer 16 closest to back- barrier layer 14, back barrier layer 14 can have a lattice constant that is within 0.5% or within 0.2% of that of the buffer layer 12. For example, when at least a substantial portion of the thickness of layer 12, for example at least half the thickness of layer 12, is GaN, layer 14 can be Al _x In _y Ga _z N, where x+y+z is about 1 and the ratio x/y is between about 81/19 and 84/16. In other implementations, layer 14 includes AlInN, where the Al composition is between 0.82 and 0.83 or between 0.81 and 0.84.

[0022] It has been found that when back barrier layer 14 has a lattice constant that is within 0.5%, or within 0.2% of that of the buffer layer 12, both the sheet charge carrier density and the carrier mobility in the 2DEG in the access regions is increased, resulting in a substantially lower sheet resistance of the 2DEG in the access regions. For example, when Al ₈₂ In ₁₈ N is used for the back barrier layer 14 and GaN is used for the buffer layer 12 and for the channel layer 16, a sheet charge carrier density of about 1.7 x 10 ¹³ cm ^"3 and a mobility of about 1 100 cm ² /V ^» s is achieved, resulting in a sheet resistance of about 334 ohms/square, which is less than 400 ohms/square. This is a substantially lower sheet resistance than has previously been reported for N-polar III-N devices. For comparison, when the back barrier layer in the structure described above was replaced with an Al _w Gai_ _w N layer with w between 0.41 and 0.61, the highest sheet charge carrier density achieved was about 1.3 x 10 ¹³ cm ^"3 , which occurred when w=0.41. The highest mobility achieved was about 524 cm ² /V ^» s, which occurred when w=0.61. The lowest sheet resistance achieved was greater than 1090 ohms/square, which also occurred when w=0.61. It was unexpected that such an improvement in the 2DEG sheet charge carrier density, mobility, and sheet resistance could be achieved by utilizing a back barrier layer with a lattice constant that is within 0.5%, or within 0.2% of that of the buffer layer.

[0023] In some implementations, an additional AIN layer or Al _x Gai_ _x N layer

(not shown), preferably with x>0.5, is included either between layers 14 and 16 or between layers 12 and 14, or both. The additional AIN or Al _x Gai_ _x N layer(s) can have a larger bandgap than the back-barrier layer 14, and may have a lattice constant which is smaller than that of buffer layer 12, in order to increase the charge density in the portion of 2DEG 19 which is in the device access regions.

[0024] In order to minimize the resistance of the channel in the device access regions (i.e., the access resistance of the device), such that resistive losses are not too large in high-voltage switching applications, the sheet resistance of the 2DEG 19 in the access regions should be kept below 400 ohms/square. The access resistance typically increases with decreasing n-type impurity concentration in layer 14, decreasing thickness of layer 14, and decreasing thickness of layer 16. For example, if layer 14 is AlInN that has a thickness of about 15 nanometers and is doped with n- type impurities at a concentration of about 4 x 10 ¹⁸ cm ^"3 , the thickness of the channel layer can be less than 15 nm, but is typically less than 10 nm. In devices that have such a low access resistance, the carrier concentration in the portion of the 2DEG 19 which is in the access regions is typically greater than 1 x 10 ¹³ cm ^"2 , and can be as high as 1.7 x 10 ¹³ cm ^"2 or higher. In previous N-polar E-mode transistors that do not include a back-barrier having a lattice constant so closely matched to that of the buffer layer 12, such low access sheet resistances could not be achieved with such a thin channel layer 16, i.e., less than 10 nanometers. Having such a thin channel layer can improve the coupling between the gate and the channel in the gate region, as well as relaxing the requirements necessary for achieving an E-mode device, as will be described below.

[0025] Still referring to Figure 1, III-N layer 18 depletes the portion of the

2DEG channel 19 in the gate region, such that when 0V is applied to the gate relative to the source, the 2DEG 19 is substantially depleted in the gate region and the device is in the off state. A positive voltage must be applied to the gate relative to the source in order to populate the 2DEG 19 in the gate region with carriers (electrons) and thereby turn the device on. The minimum gate voltage (relative to the source) required to turn the device on is known as the device threshold voltage.

[0026] In order for the portion of 2DEG channel 19 in the gate region to be depleted of mobile charge when 0V is applied to the gate 23 relative to the source 21, layer 18 is sufficiently thick, for example thicker than 10 nanometers, and typically has a bandgap that is greater than that of channel layer 14, which can be achieved by ensuring that layer 18 has a sufficiently large aluminum fractional composition. To enable a sufficiently thick gate-barrier layer 18, the gate barrier layer 18 can be formed of a III-N material having a lattice constant that is within 0.5% or within 0.2% of that of the buffer layer 12. For example, when at least a substantial portion of the thickness of layer 12 (i.e., at least half the thickness of layer 12) is GaN, layer 18 can be Al _x In _y Ga _z N, where x+y+z is about 1 and the ratio x/y is between about 81/19 and 84/16. In some implementations, layer 18 includes AlInN (i.e., Al _x Ini_ _x N), where the Al composition x is between 0.82 and 0.83.

[0027] In many cases, layer 18 is unintentionally doped n-type, for example as a result of impurities being incorporated into the layer during growth. Layer 18 can therefore include p-type dopants to counterbalance the incorporated n-type impurities and further increase the threshold voltage of the device. In cases where layer 18 is substantially depleted of mobile charge carriers, the device can be formed either with or without insulator layer 20. However, if the p-type dopants induce a substantial concentration of holes in the gate-barrier layer 18, the device may be rendered inoperable if the insulator layer 20 is included. In this case, the device is formed without an insulator layer 20 between gate 23 and layer 18, or with gate 23 contacting at least a portion of the surface of layer 18 closest to the gate 23.

[0028] As seen in Figure 2, in some implementations layer 18 extends all the way from source 21 to drain 22, or at least a portion of layer 18 is over the portion of 2DEG channel 19 in the access regions. In order to ensure that the portion of the 2DEG channel 19 in the access regions is not depleted of mobile charge, which would render the device inoperable, the portion of layer 18 in the access regions is made thinner than the portion beneath the gate. Alternatively, the portion of layer 18 in the access regions could have a lower p-type doping concentration than the portion of layer 18 underneath the gate.

[0029] Another N-polar Ill-Nitride E-mode transistor 300 is shown in Figures

3A and 3B. Figure 3A illustrates the transistor 300 while 0V is applied to the gate relative to the source, such that the transistor is biased OFF, and Figure 3B illustrates the transistor 300 when a positive voltage greater than the threshold voltage of the transistor is applied to the gate relative to the source, such that the transistor is biased ON. Similarly to the transistors of Figures 1 and 2, transistor 300 optionally includes a substrate 10, and also includes a III-N buffer layer 12 and a first III-N barrier layer 14 on the III-N buffer layer 12. A first III-N channel layer 26 is on the first III-N barrier layer 14, and a second III-N barrier layer 28 is on the first III-N channel layer 26. Transistor 300 also includes a second III-N channel layer 36 on the second III-N barrier layer 28, and optionally includes a III-N cap layer 32 on the second III-N channel layer 36. An aperture or recess 37 is formed in the III-N channel layer 36 over the gate region of the transistor. In Figure 3 A, layers 30 and 33 already have been formed in this aperture or recess. The aperture or recess 37 is shown to extend through the entire thickness of the III-N channel layer 36, but it may optionally extend only partially through the III-N channel layer 36 (not shown). When the III-N cap layer 32 is included, the aperture or recess 37 is formed through the entire thickness of the III-N cap layer 32, as shown, before layer 30 and gate 33 are formed. After forming the recess or aperture 37, insulator layer 30 is deposited conformally over the exposed upper surface of III-N material, including in the recess or aperture 37, as shown. Gate 33 is then formed over the insulator layer and is at least partially in the recess or aperture 37. Source and drain contacts 21 and 22, respectively, are on opposite sides of the gate 33. In some implementations, insulator layer 30 is not included, and gate 33 is formed directly on the III-N material layers (not shown).

[0030] The compositions and thicknesses of the layers in transistor 300 are selected as follows. As in the transistors of Figures 1 and 2, the III-N buffer layer 12 can be substantially thicker than any of the other Ill-Nitride layers in the structure. For example, layer 12 may have a thickness that is at least 10 times, but typically at least 30 times, the combined thickness of the other III-N layers. In some

implementations, the III-N buffer layer includes or is formed of GaN. In other implementations, the upper portion of the III-N buffer layer 12 (i.e., the portion of layer 12 closest to layer 14) is GaN. When the III-N buffer layer 12 is substantially thicker than the other Ill-Nitride layers in the structure, the other III-N layers are strained unless their lattice constants are about the same as that of the III-N buffer layer 12.

[0031] The first III-N barrier layer 14 may have a lattice constant that is substantially the same as that of III-N buffer layer 12, such that the first III-N barrier layer 14 is under little or no strain. That is, the lattice constant of the first III-N barrier layer 14 may be within 0.5% or within 0.2% of the lattice constant of the III-N buffer layer 12. The first III-N barrier layer 14 also has a bandgap which is larger than that of the first III-N channel layer 26. For example, when the first III-N channel layer 26 and the upper portion of the III-N buffer layer 12 are GaN, the first III-N barrier layer 14 can be formed of or include AlInN (i.e., Al _x Ini_ _x N), where the Al composition x is between 0.82 and 0.83 or between 0.81 and 0.84. Alternatively, when the first III-N channel layer 26 and the upper portion of the III-N buffer layer 12 are GaN, the first III-N barrier layer 14 can be formed of or include Al _x In _y Ga _z N, where x+y+z is about 1 and the ratio x/y is between about 81/19 and 84/16. In some implementations, an additional AIN layer or Al _x Gai_ _x N layer (not shown), preferably with x>0.5, is included either between layers 14 and 26 or between layers 12 and 14, or both. The additional AIN or Al _x Gai_ _x N layer(s) can have a larger bandgap than that of the first III-N barrier layer 14, and may have a lattice constant which is smaller than that of III-N buffer layer 12, in order to increase the charge density in the second 2DEG 39 (shown in Figure 3B) which is induced in the first III-N channel layer 26 when the gate of transistor 300 is biased on, as further described below.

[0032] The second III-N barrier layer 28 has a bandgap which is larger than that of both the first III-N channel layer 26 and the second III-N channel layer 36, and the composition of layer 28 can be selected such that its lattice constant is substantially the same as that of III-N buffer layer 12, so that the second III-N barrier layer 28 is under little or no strain. That is, the lattice constant of the second III-N barrier layer 28 may be within 0.5% or within 0.2% of the lattice constant of the III-N buffer layer 12. For example, when the first and second III-N channel layers 26 and 36, as well as the upper portion of the III-N buffer layer 12, are all GaN, the second III-N barrier layer 28 can be formed of or include AlInN (i.e., Al _x Ini_ _x N), where the Al composition x is between 0.82 and 0.83 or between 0.81 and 0.84. Alternatively, the second III-N barrier layer 28 can be formed of or include Al _x In _y Ga _z N, where x+y+z is about 1 and the ratio x/y is between about 81/19 and 84/16. In another implementation, the second III-N barrier layer 28 can be formed of or include A1N or AlGaN, having a larger bandgap than both layers 26 and 36, but having a smaller lattice constant than the III-N buffer layer 12.

[0033] The thickness of the second III-N barrier layer 28 is selected to be large enough that the second 2DEG 39 is substantially depleted of charge, or mobile charge is not induced in the gate region of the first III-N channel layer 26, when 0 Volts is applied to the gate 33 relative to the source 21, as illustrated in Figure 3A. That is, during the fabrication of transistor 300, immediately after the formation of layer 26 and before the formation of any of the layers overlying layer 26, a 2DEG channel (not shown) is induced in layer 26 adjacent to the interface between layers 26 and 14. As layer 28 is deposited over layer 26, the charge concentration in the induced 2DEG gradually decreases as layer 28 is made thicker. There exists a minimum thickness for layer 28, the minimum thickness depending on the compositions of layers 26 and 28 as well as the thickness of layer 26, such that when layer 28 is grown thicker than this minimum thickness, the induced 2DEG becomes substantially depleted of charge and hence ceases to be conductive. In transistor 300, layer 28 is made thicker than this minimum thickness, thereby ensuring that 2DEG 39 is induced in the gate region of the first III-N channel layer 26 when a positive voltage greater than the transistor threshold voltage is applied to the gate 33 relative to the source 21, but not when 0V is applied to the gate relative to the source. For example, when the first III-N channel layer 26 is GaN and is about 5 nanometers or thinner, and the second III-N barrier layer 28 is Al _w In _1-W N with w approximately in the range of 0.82 to 0.83, layer 28 can be at least 2 nanometers, between 2 and 10 nanometers, or between 2 and 6 nanometers.

[0034] The second III-N channel layer 36, which can for example be formed of or include GaN, has a smaller bandgap than the second III-N barrier layer 28, such that 2DEG channel 29 is induced in layer 36 both when the gate of transistor 300 is biased ON and when the gate of transistor 300 is biased OFF. For a given composition of layers 28 and 36, the mobile charge density in 2DEG 29 depends on the thickness of layer 36. Increasing the thickness of layer 36 tends to increase the mobile charge carrier density in 2DEG channel 29. Having a higher 2DEG mobile carrier density can decrease the sheet resistance in the access regions of the transistor 300, leading to lower conductive losses in the transistor during operation. However, a higher 2DEG mobile carrier density also results in a lower breakdown voltage of the device. Hence, when the transistor 300 is used in applications that require a minimum transistor breakdown voltage, the thickness of layer 36 can be selected such that the resulting 2DEG mobile carrier density is about equal to or just below the carrier density that results in the transistor having the required minimum breakdown voltage.

[0035] In some implementations, an additional AIN layer or Al _x Gai_ _x N layer

(not shown), preferably with x>0.5, is included either between layers 28 and 26 or between layers 28 and 36, or both. The additional AIN or Al _x Gai_ _x N layer(s) can have a larger bandgap than that of the second III-N barrier layer 28, and may have a lattice constant which is smaller than that of III-N buffer layer 12, in order to increase the charge density and/or the mobility of the 2DEG 29 which is induced in the second III- N channel layer 36 in the access regions of the transistor 300.

[0036] The III-N cap layer 32, when included in transistor 300, can have a larger bandgap than the second III-N channel layer 36, which in some

implementations increases the breakdown voltage of the transistor. Alternatively, the cap layer 32 may have a bandgap which is smaller than or similar to that of layer 36, provided the conduction band edge of the cap layer 32 at the interface between layers 32 and 36 is at a higher absolute energy than the conduction band edge of the layer 36 at the interface between layers 32 and 36 III-N cap layer 32 may also be designed to have a similar lattice constant to that of III-N buffer layer 12. For example, when layers 12 and 36 are both GaN, the III-N cap layer 32 can be formed of or include AlInN (i.e., Al _x Ini_ _x N), where the Al composition x is between 0.82 and 0.83 or between 0.81 and 0.84. Alternatively, the III-N cap layer 32 can be formed of or include Al _x In _y Ga _z N, where x+y+z is about 1 and the ratio x/y is between about 81/19 and 84/16. Including the III-N cap layer 32 increases the separation between the uppermost III-N surface and the 2DEG 29 in the access regions of the transistor 300, which can reduce dispersion in the transistor during operation.

[0037] Insulator layer 30, which can for example be formed of or include silicon nitride, aluminum oxide, aluminum nitride, silicon dioxide, or various combinations of these insulator materials, serves as a gate insulator layer, which reduces gate leakage in the device. Insulator layer 30 can also serve as a surface passivation layer in the access regions of the transistor 300, in order to minimize or reduce dispersion transistor during operation.

[0038] The transistor 300 of Figures 3 A and 3B operates as follows. As seen in Figure 3A, when the transistor 300 is biased OFF, for example with 0 Volts applied to the gate 33 relative to the source 21, conductive 2DEG channel 29 is in the access regions of the transistor 300 in layer 36, but no conductive channel is present in the gate region of the transistor. Referring to Figure 3B, when the transistor 300 is biased ON, for example by applying a sufficiently positive voltage to the gate 33 relative to the source 21, 2DEG channel 39 is induced in layer 26 in the gate region of the device. The positive gate voltage can also induce conductive portions 49, which electrically connect 2DEGs 29 and 39, thereby forming a continuous electrically conductive path extending from the source 21 to the drain 22.

[0039] Still referring to Figure 3B, having the portion of the channel in the access regions be in a different layer from the portion of the channel in the gate region can be advantageous in that the design parameters for the gate and access regions of the device can be decoupled from one another. That is, in the transistors of Figures 1 and 2, designing the device to have a higher mobile carrier charge density in the access regions of the transistor reduces the maximum threshold voltage that can be obtained for the transistor. On the other hand, for transistor 300 of Figures 3A and 3B, the transistor threshold voltage and the mobile carrier charge density in the access regions of the transistor can be controlled independently of one another. For example, the thickness of layers 26 and 28 can be selected to produce the desired threshold voltage of the transistor, and the thickness of layer 36 can then be selected to produce the desired charge density in 2DEG 29 in the device access regions, since varying the thickness of layer 36 does not substantially vary the transistor threshold voltage. For example, layers 26 and 28 can each be less than 6 nanometers thick, such as about 3 nanometers thick, while layer 36 can be greater than 10 nanometers thick, such as about 15 nanometers thick. Having the device channel in the access regions of the device be in a different layer from the device channel in the gate region of the device, and in particular having layer 36 (which contains the 2DEG in the device access regions) be at least 2 times or at least 3 times as thick as layer 26 (which contains the 2DEG in the gate region of the device), can result in an N-Polar enhancement-mode Ill-Nitride transistor having a breakdown voltage greater than 300 Volts or greater than 600 Volts, a channel sheet resistance of less than 400 ohms/square in the access regions of the device while maintaining a threshold voltage of at least 2 Volts.

Furthermore, because the source and drain contacts 21 and 22, respectively, do not need to be formed through a barrier layer, the source and drain contact resistances can be less than 2 ohm-mm.

[0040] A number of implementations have been described. Nevertheless, it will be understood that various modifications may be made without departing from the spirit and scope of the techniques and devices described herein. For example, an additional A1N layer or Al _z Gai_ _z N layer (not shown) with z>0.5 or z>0.8 can be included between layers 18 and 16 in either device of Figure 1 or Figure 2. The additional A1N or Al _x Gai_ _x N layer can have a larger bandgap than the gate-barrier layer 18, and may have a lattice constant which is smaller than that of buffer layer 12, in order to increase the charge density in the 2DEG 19. Or, the portion of the buffer layer 12 which is closest to barrier layer 14 can be undoped or unintentionally doped and can be formed of GaN, and optionally an n-type III-N layer such as n-type GaN can be included between the buffer layer 12 and the barrier layer 14. Furthermore, III-N barrier layer 14 or 28 can be made very thin, and can for example be a finite thickness that is less than 1 nanometer, for example about 0.5 nanometers or less. Having such a thin barrier layer can result in a larger mobility for a given charge density in the 2DEG in the transistor access regions. As such, a given sheet resistance in the transistor access regions can be achieved with a lower 2DEG charge density in the access regions, thereby resulting in a larger breakdown voltage. Additionally, in transistor 300 of Figures 3A and 3B, an additional p-type III-N layer (not shown) can be included between the gate 37 and the underlying III-N layer 26 or 28 in the recess or aperture 37. When the additional p-type III-N layer is included, the gate 33 can directly contact the p-type III-N layer. In this case, the portion of layer 30 in the gate region can be omitted, such that layer 30 is over the device access regions but is not in the gate region of the transistor. Accordingly, the scope of the disclosure is governed by the following claims.