Abstract
GaNbased highelectron mobility transistors (HEMTs) with AlN/GaN superlattices (SLs) (4 to 10 periods) as barriers were prepared on (0001) sapphire substrates. An innovative method of calculating the concentration of twodimensional electron gas (2DEG) was brought up when AlN/GaN SLs were used as barriers. With this method, the energy band structure of AlN/GaN SLs was analyzed, and it was found that the concentration of 2DEG is related to the thickness of AlN barrier and the thickness of the period; however, it is independent of the total thickness of the AlN/GaN SLs. In addition, we consider that the sheet carrier concentration in every SL period is equivalent and the 2DEG concentration measured by Hall effect is the average value in one SL period. The calculation result fitted well with the experimental data. So, we proposed that our method can be conveniently applied to calculate the 2DEG concentration of HEMT with the AlN/GaN SL barrier.
Introduction
The GaNbased highelectron mobility transistor (HEMT) is a promising research subject because of the expected advantages for the realization of electronic devices for highpower and hightemperature operations [15]. There are also applications as biosensors for detection of bacteria, DNA, and so on. The Al_{0.3}Ga_{0.7}N/AlN/GaN structures achieved a twodimensional electron gas (2DEG) mobility of 2,185 cm^{2}/V·s at room temperature with the carrier density of 1.1 × 10^{13 }cm^{2 }which was deposited by metalorganic chemical vapor deposition (MOCVD) on sapphire substrate [6]. However, the relatively low carrier sheet density with AlGaN/GaN structures hinders the further development of device process. Although the sheet carrier density can be further improved by increasing the Al content in the ternary barrier layer, this increases the strength of polarization; on the other hand, increasing the Al content in the ternary layer may deteriorate the quality of AlGaN due to large lattice mismatch, and ternary alloy scattering in the heterostructure resulted in poor transport properties [7]. In contrast, the growth of AlN barrier layer on GaN channel enhances the transport properties. Introducing nanostructures into the HEMT epilayers can enhance the device performance. For example, recent reports have shown that by replacing the conventional AlGaN barrier layers with AlN/GaN superlattices composed of severalnanometerthick binary alloys, electric field induced by macroscopic polarization is much stronger, and the sheet carrier density is higher [8]. The AlN/GaN SLs have stronger polarization effect and larger conduction band offset than the heterostructure of AlGaN/GaN, and the SLs have deeper triangular quantum well.
The carriers in the heterointerface are confined twodimensionally and have stronger quantum effect than heterostructure of AlGaN/GaN as sheet carrier density, which makes more density of 2DEG and higher mobility in GaNbased HEMT device possible. Despite excellent transport properties, the described AlN/GaN structures have an obvious drawback: the 2DEG electron channel lies very close to the surface which makes it very sensitive to any process occurring at the surface of the sample. Compared with AlN as barrier, the 2DEG electron of HEMT with the AlN/GaN SL barrier has strong antijamming because it has several similar 2DEG electron channels which reduce any influence from surface by several times, and the sheet carrier density of HEMT with the SL barrier is much higher than AlGaN as barrier.
The purpose of this paper is to present detailed calculations about HEMT with the SLs as the barrier layers. There are many studies on growing the HEMT with SL barrier but there are few reports on calculations about the sheet carrier density of heterostructure with SLs as barrier. As the energy band of HEMT with SL barrier is analyzed, we assume that the 2DEG will be formed in every SL period if the thickness of AlN is beyond the critical thickness (d_{0}) which is the least limit thickness of AlN barrier to form the 2DEG. We also consider that the 2DEG density in each SL period is equal since the SLs have periodicity and tunnel impenetration. We deduce the equation which can be applied to calculate the 2DEG density in GaN/AlN/GaN heterostructures and calculated the 2DEG density with the data of our SLs, and at the same time, the calculated results were compared with experimental data.
Experimental details
AlGaN/GaN heterostructures were grown by MOCVD on (0001) sapphire substrates using AlN/GaN SL (4 to 10 periods) structures as barrier layers where the topmost layer was GaN. Trimethylgallium, trimethylaluminum, and ammonia were used as Ga, Al and, N precursors, respectively. The schematic diagram is shown in Figure 1. Growth conditions of semiinsulated GaN buffer of all samples were the same and they had no intended doping. The equivalent Al composition of the quasialloy can be changed by adjusting the thicknesses of AlN and GaN layers. Highresolution Xray diffraction was performed to measure the thickness of AlN/GaN SLs of the samples using Bede D1 system. The equivalent Al composition is defined as Equation 1:
Figure 1. Schematic diagram of GaN/AlN HEMT. Area shaded in light and dark grey stripes, GaN/AlN Sls; light grey, GaN; dark grey, HTAlN; red, sapphire.
As shown in Figure 2, the simulation line is fitting well with the experiment line so we can exactly determine the thicknesses of AlN and GaN. The satellite peaks which are related with the diffraction of SLs can be seen in each picture indicating that the period of SLs is uniform and the interface is smooth.
Figure 2. The ω2θ scans of SL structures. (a) and (b) are the Images of SL structures with different Al compositions and periods. Red lines, experimental values; black lines, simulation values.
Three groups of heterostructures were grown. One group consisted of different Al compositions, and the other two consisted of equivalent Al compositions. While the SLs in the second group had different period thicknesses, the last group had the same period (3.5 nm) but with different numbers of SLs (4 to 6 periods). Hall effect measurements were performed using the vander Pauw method on cleaved 6mm × 6mm squares using In/Ga alloy as ohmic contacts.
Results and discussion
In the first group of heterostructures, the nominal equivalent Al composition of the quasialloy varies from 22% to 40%. Figure 3a shows that sheet carrier density increased rapidly as the nominal equivalent Al composition increased. In the second group, the nominal Al compositions of the quasialloy are all 25%, and the sheet carrier density as a function of the SL period thickness is plotted in Figure 3b. The sheet carrier densities are increasing, with the SL period getting thick. In the third group, the AlN/GaN SL barrier had the same Al composition (Al% = 36.5%) and thickness of one SL period. As shown in Figure 3c, the sheet carrier densities are about the same when the periods of SLs increase. In other words, when the AlN/GaN SLs are under full strain, the sheet carrier densities are independent of the total thickness of AlN/GaN SLs. This is different from the situations in traditional AlGaN barrier HEMTs where the sheet carrier density is influenced by the Al composition and thickness of AlGaN. In the case of HEMT with the AlN/GaN SL barrier, we find that the sheet carrier densities are increasing with the period thickening of SLs, and yet, Al composition is still independent of the total thickness of AlN/GaN SLs.
Figure 3. Relations of sheet carrier concentrations. Relations with (a) Al mole fraction, (b) SL period thickness, and (c) SL periods. Black filled squares, experimental values; red diamonds, theoretical values.
In traditional AlGaN barrier HEMT structures, we can calculate the 2DEG density using Equations 2 and 3 [9,10]:
where ε_{0 }is vacuum permittivity, n_{s }is the sheet carrier density of HEMT, d_{0 }is the critical thickness of AlN which can form 2DEG. If the thickness of AlN is smaller than d_{0}, it could not form the 2DEG. ε_{AlGaN }is the relative dielectric constant of AlGaN, d_{AlGaN }is the thickness of the barrier, eφ_{B }is the Schottky barriers of the gate contact on top of AlGaN, E_{F}(x) is the Fermi level with respect to the GaN conductionbandedge energy, and ΔE_{C }is the conduction band offset at the AlGaN/GaN interface where 2DEG forms.
When AlN is deposited on GaN, it will produce negative (acceptorlike) piezoelectric polarization (PZ) charge density at the top of the AlN barrier layer and positive PZ charge density at the AlN/GaN interface [8] as shown in Figure 4 (a). There is no 2DEG until the thickness of AlN barrier is beyond d_{0 }which can be determined by Equation 2.
Figure 4. The 2DEG between AlN/GaN interface with different AlN thickness. (a) The thickness of AlN barrier less than the critical thickness (d_{0}), (b) the thickness of AlN barrier beyond d_{0}, and (c) the thickness of AlN barrier beyond d_{0}, and the GaN was deposited on the AlN barrier.
The energy E_{D }of surface state is below the E_{F }if the thickness of AlN is less than d_{0}. With barrier thickness increasing, E_{F}E_{D }is decreasing. In addition, the energy E_{1}, the surface energy of conduction band, increases with increasing barrier thickness because there is a constant electric field in the AlN barrier due to the unscreened polarization field. At d_{0 }(E_{F}E_{D }= 0), the donors at a free surface will begin to be ionized, and the PZ charge density at the AlN/GaN interface will start to be compensated by the 2DEG formed at the interface. The charge at the top of the AlN will be compensated by charged surface states for a free surface when the thickness of AlN barrier is beyond the critical thickness. At critical thickness (E_{F}E_{D }= 0), electrons are able to transfer from occupied surface state to empty conduction band states at the interface. The 2DEG density is increasing with increasing barrier thickness, while the E_{D }and E_{1 }are constant and independent on the barrier thickness. The E_{1 }is dependent on the material of the barrier layer.
Base on these analyses, the interfacial charge between AlN and GaN that determines the density of 2DEG is mainly dependent on the polarized intensity. For this reason, we calculate the 2DEG density with AlN barrier according to Equation 2, and the result is shown in Equation 3 [911]:
where and are the spontaneous polarization of AlN and GaN, and is the piezoelectric polarization of AlN buffer layer; is the polarizationinduced charge density determined by the difference in the total spontaneous and piezoelectric polarization within AlN and GaN layers; n_{s }is the sheet carrier density; and d_{0 }is the critical thickness of AlN which can form 2DEG.
The 2DEG could also influence E_{F}; however, according to Equation 4 [12], we find that the change of E_{F }is less than 10% even if the n_{s }has a change of 10^{12 }when the sheet carrier density is up to 10^{13 }cm^{2}. In order to simplify the computing, we ignore the variation of E_{F }coursed by 2DEG:
However, when GaN is deposited on the AlN barrier, it can change the energy band as shown in Figure 4 (c). The surface state is pined on the energy E_{F }because the thickness of AlN barrier is beyond d_{0}. The energy E_{1 }rises with the increasing thickness of GaN since the electric field also rises from the 2DEG has not been screened. According to the equilibrium of total electric field, we take a simple electrostatic analysis on the energy band and gain the result as shown in Equations 5 and 6:
By speculating the equation, the density of 2DEG can be expressed as:
where d_{1 }and d_{2 }are the thicknesses of AlN barrier and GaN layer, respectively; E_{F }is the Fermi level position with respect to the GaN conductionband edge; and ε_{1 }and ε_{2 }are the dielectric constants of AlN and GaN, respectively. According to Equation 6, the 2DEG density is dependent on both the Al mole fraction and the thickness of SL period but independent on the total thickness of SLs when the AlN/GaN SLs are used as the barrier layer.
In fact, the sheet carrier density measured using Hall coefficient is the average value of all SLs. It is difficult to calculate the 2DEG density in every SL period directly. The sandwich structure shown in Figure 4 (c) can be seen as one period in SL barrier. We consider that the 2DEG will be formed in the every SL period, and the sheet carrier density is equal in every SL period since the SLs have periodicity and tunnel impenetration. So, the n_{s }in each channel could be calculated using Equation 6. Therefore, we think that we can simply calculate 2DEG density of the outermost SL period simply to represent the 2DEG density in every SL.
As is shown in Figure 3, the red diamonds correspond to simulate value by means of Equation 6 where the surface barrier value e φ_{B }is fixed at 0.55 eV, which was obtained by fitting the first experimental data points in Figure 3a. Although there was no report about the eφ_{B }of In/Ga alloy on GaN as seen in Figure 3a, b, c, we can see that the simulate values are accordingly well with the experimental value; hence, we think that Equation 6 can be used to compute the 2DEG density.
However, as shown in Figure 3b, the simulate values highly contradict with the experimental values. If the thickness of SL is small, the energy band bending in each SL period should not be the same, and the sheet carrier density is not equal in every SL period. As the interface of heterostructure is about 5Å thick, the interfacial influence may become notable as the thickness of one SL period decreases which makes it more difficult to compute for the 2DEG density.
Conclusions
AlN/GaN SLs as the barrier of HEMT were grown on semiinsulated GaN, and the formation of 2DEG was researched particularly. The 2DEG concentration, characterized by Hall effect measurements, was found to be as a function of AlN thickness in SL period but independent on the total thickness of SLs. The calculation with our innovative method have proven these laws. During the process of calculation, we consider that the 2DEG would form in every SL period when the AlN barrier thickness in SL period was beyond the critical thickness, and concentration of the 2DEG was equivalent in each period. However, once the AlN barrier thickness in SL period is smaller than the critical thickness, the 2DEG would not be formed in any SL period according to the theory. Besides, the interfacial influence may become notable, which makes it much more difficult to calculate the density. The model we set has proved to be fitted well with the experimental result and can be used to calculate the 2DEG density in HEMTs with SL barriers.
Abbreviations
2DEG: twodimensional electron gas; HEMT: highelectron mobility transistor; MOCVD: metalorganic chemical vapor deposition; PZ: piezoelectric polarization; SLs: superlattices.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
PX and YC contributed the main ideas of superlattice structure design and the calculation of 2DEG density. YJ, ZM, XW, ZD, and YL carried out the MOCVD growth, the Xray diffraction, and the Hall effect measurements. HJ, WW, and HC gave important advices to the paper. All authors read and approved the final manuscript.
Acknowledgements
Present research work was supported by the National High Technology Research and Development Program of China (grant nos. 2011AA03A112 and 2011AA03A106) and the National Nature Science Foundations (grant nos. 60890192, 50872146, and 60877006).
References

Khan MA, Chen Q, Shur MS, McDermott BT, Higgins JA, Burm J, Schaff WJ, Eastman LF: Microwave operation of GaN/AlGaNdoped channel heterostructure field effect transistors.

Wu YF, Keller BP, Keller S, Kapolnek D, Kozodoy P, DenBaars SP, Mishra UK: Very high breakdown voltage and large transconductance realized on GaN heterojunction field effect transistors.
Appl Phys Lett 1996, 69:14381440. Publisher Full Text

Gaska R, Yang JW, Osinsky A, Chen Q, Asif Khan M, Orlov AO, Snider GL, Shur MS: Electron transport in AlGaNGaN heterostructures grown on 6HSiC substrates.
Appl Phys Lett 1998, 72:707709. Publisher Full Text

Ozgur A, Kim W, Fan Z, Botchkarev A, Salvador A, Mohammad SN, Sverdlov B, Morkoc H: High transconductance normallyoff GaN MODFETs.
Electron Lett 1995, 31:13891390. Publisher Full Text

Binari SC, Redwing JM, Kelner G, Kruppa W: AlGaN/GaN HEMTs grown on SiC substrates.
Electron Lett 1997, 33:242243. Publisher Full Text

Wang X, Wang C, Hu G, Xiao H, Fang C, Wang J, Ran J, Li J, Li J, Wang Z: MOCVDgrown highmobility Al_{0.3}Ga_{0.7}N/AlN/GaN HEMT structure on sapphire substrate.

Wang C, Wang X, Hu G, Wang J, Li J: Influence of Al content on electrical and structural properties of Sidoped Al_{x}Ga_{1x}N/GaN HEMT structures.
Phys Stat Sol 2006, 3:486489. Publisher Full Text

Smorchkova IP, Chen L, Mates T, Shen L, Heikman S, Moran B, Keller S, DenBaars SP, Speck JS, Mishra UK: AlN/GaN and (Al, Ga)N/AlN/GaN twodimensional electron gas structures grown by plasmaassisted molecularbeam epitaxy.
J Appl Phys 2001, 90:51965201. Publisher Full Text

Chu M, Koehler AD, Gupta A, Nishida T, Thompson SE: Simulation of AlGaN/GaN highelectronmobility transistor gauge factor based on twodimensional electron gas density and electron mobility.
J Appl Phys 2010, 108:104502. Publisher Full Text

YuanJie L, ZhaoJun L, Yu Z, LingGuo M, ZhiFang C, ChongBiao L, Hong C, ZhanGuo W: Influence of thermal stress on the characteristic parameters of AlGaN/GaN heterostructure Schottky contact.
Chin Phys B 2011, 20:047105. Publisher Full Text

Kozodoy P, Hansen M, DenBaars SP, Mishra UK: Enhanced Mg doping efficiency in Al_{0.2}Ga_{0.8}N/GaN superlattices.
Appl Phys Lett 1999, 74:36813683. Publisher Full Text

Ambacher O, Smart J, Shealy JR, Weimann NG, Chu K, Murphy M, Schaff WJ, Eastman LF, Dimitrov R, Wittmer L, Stutzmann M, Rieger W, Hilsenbeck J: Two dimensional electron gases induced by spontaneous and piezoelectric polarization in undoped and doped AlGaN/GaN heterostructures.
J Appl Phys 1999, 85:32223233. Publisher Full Text