掺杂维度和浓度调控的d掺杂的La:SrTiO3超晶格结构金属-绝缘体转变*李云1)† 鲁文建2)1) (韩山师范学院物理与电子工程学院, 潮州 521041)2) (中国科学院合肥物质科学研究院固体物理研究所, 合肥 230031)(2021 年4 月30日收到; 2021 年7 月3日收到修改稿)利用密度泛函理论计算, 本文系统研究了d掺杂的La:SrTiO3超晶格结构的电子性质随掺杂维度和掺杂浓度改变而变化的规律性. 该结构通过在SrTiO3等间距的单元层中掺入一定浓度的La来实现. 在25% La掺杂浓度下, 随着相邻掺杂层间距从1个单层增加到5个单层, 掺杂维度从三维过渡到二维, 超晶格从金属性变到绝缘体性, 并在带隙中产生局域态, 且该局域态呈现出电荷序、自旋序和轨道序. 这种金属-绝缘体转变是由于二维电子体系呈现出更强的关联性造成的. 而随着二维掺杂浓度提高到50%, 关联性降低, 体系变成金属性.关键词：二维掺杂, 强关联, 金属-绝缘体转变PACS：71.30.+h, 73.20.–r, 71.27.+a DOI: 10.7498/aps.70.202108301 引 言由于具有较强的电子关联性, 过渡金属氧化物经常展现出一些非凡的特性, 其晶格、电子和磁性构型与一些引人注目的物理性质存在着紧密的相关性[1]. 这些特性通常是电子电荷、自旋和轨道自由度与晶格微妙作用的结果[2,3]. 调控这些参数有可能产生丰富的电学性质, 有助于发展出有新特性的功能材料和器件[4,5]. 比如, 掺杂能调控材料内部的多个自由度, 如电荷、自旋、轨道占据等, 通过掺杂改变d轨道的填充度能诱导过渡金属氧化物实现金属-绝缘体转变、反铁磁-铁磁转变、普通导体-超导体转变等[1−3,5]. 这种调控手段和对应的性能已经在传感器、自旋电子器件、存储器等领域得到应用.电子体系的维度也是影响材料内部电子学特性的一个重要调控参数. 随着电子体系的维度降低, 如从三维变到二维, 电子间的关联作用变强并可能起主导作用[6]. 这可能导致电子体系呈现出整体有序性并使得体系进入非同于三维体系的新的相. 随着材料生长技术的进步, 制备二维电子体系已经比较容易实现, 如在异质界面体系或者二维掺杂体系. 最近几年在钙钛矿过渡金属氧化物界面体系的研究表明过渡金属氧化物异质结二维电子体系具备一些独特的性质[7−21], 如金属-绝缘体转变、磁性-非磁性转变、超导等. 为了研究过渡金属氧化物电学性质随着掺杂维度和掺杂浓度改变而变化的规律, 本文利用密度泛函理论计算研究了d掺杂(在一个单元层中掺杂La而其近邻层不掺杂)的La:SrTiO3超晶格结构的电子性质, 通过改变掺杂层的间距可实现从三维掺杂过渡到二维掺杂, 并通过改变掺杂浓度来改变二维电子的密度.* 教育部留学回国人员科研启动基金(批准号: [2015]-1098)资助的课题.† 通信作者. E-mail: liyunphy@© 2021 中国物理学会 Chinese Physical Society 计算结果表明, 调节这些参数可改变电子关联强度进而实现体系的金属-绝缘体转变.2 计算方法计算由VASP 程序包执行[22], 其中采用PBE 型广义梯度近似泛函(PBE-GGA)[23]和投影缀加平面波方法[24,25], 平面波截断动能为500 eV. Ti 3d 轨道局域性较强, 轨道中电子的在位库伦相互作用较强, 计算中采用Dudarev 的LSDA+U 方法近似描述[26]. 计算中Ti 3d 轨道电子的在位库伦相互作用能分别取U =0, 2.0, 3.0, 3.5, 3.7, 4.4, 5.0 eV 等数值, 将不同数值得到的基态电子态与实验测得的电子性质对比进而确定出合适的U 值. 图1展示了两种超晶格原胞结构, 即[Sr 0.75La 0.25TiO 3]1|[SrTiO 3]n (n = 1, 5)(简写为[SLTO]1|[STO]n ), 其中掺杂层中25%的Sr 原子被La 原子替代, 沿着[001]方向周期性重复, 面内周期为4 × 4. 相应地,采用4 × 4 × 4和4 × 4 × 2的Monkhost 型k 点网格在布里渊区中取样. 有限温度展宽采用Gaussian 方法, 其中s = 0.1 eV. 计算中所有原子都充分弛豫, 直到受力小于0.01 eV/Å.(a)(b)(c)图 1 (a) 超晶格结构面内4 × 4周期俯视图; (b) [SLTO]1|[STO]1侧视图; (c) [SLTO]1|[STO]5侧视图. 绿色球代表Sr 原子, 蓝色代表La 原子, 红色代表O 原子, Ti 原子在八面体中心Fig. 1. (a) Top view of the superlattices with in-plane 4 × 4unit cells; (b) side view of [SLTO]1|[STO]1; (c) side view of [SLTO]1|[STO]5. Green balls represent Sr atom, blue balls La atom, red balls O atom, Ti atoms are at the centre ofthe octahedrons.3 结果和讨论SrTiO 3导带底部态主要由Ti 3dt 2g (d xy , d yz ,d xz ) 轨道构成, 掺杂La 的价电子轨道5d6s 能级高于SrTiO 3中的Ti 3dt 2g 轨道能级, La 掺杂产生的电子全部进入Ti 3dt 2g 轨道能级. 计算中Ti 3d 轨道在位库伦相互作用能U 为可调参数, 本文通过比较计算结果与实验结果来确定U 的最佳数值.图2展示了两种典型的U 计算的结果. 如图2(a)和图2(b)所示, 在U = 2 eV 情况下, 费米能级穿过导带下部, 两种结构都为金属态. 在U = 3.7 eV 情况下, 如图2(c)和图2(d)所示, [SLTO]1|[STO]1仍然为金属态, 而[SLTO]1|[STO]5为绝缘态, 费米能级穿过带隙, 且在带隙里出现局域态. 计算表明当U < 3.5 eV 时两种体系都是金属态, 而当U ≥ 3.5 eV 时[SLTO]1|[STO]5才会展现为绝缘态基态. 实验中观察到[SLTO]1|[STO]1呈现金属性,而[SLTO]1|[STO]5的电阻温度曲线为绝缘态且光电导检测表明带隙内存在局域态. 又考虑了50%La 掺杂结果和带隙宽度等因素后, 确定在上述超晶格体系中U = 3.7 eV 的计算结果与实验结果吻合最好.为了澄清[SLTO]1|[STO]5带隙内局域态的性质, 图3(a)和图3(b)详细地展示了其能带结构和局域态对应的空间电荷分布. 带隙内的局域态出现在掺杂的SrO 层两侧的TiO 2层内, 掺杂电子局域在Ti 原子的3dt 2g 轨道内, 则这部分有局域电子占据的Ti 原子呈现+3价, 其他Ti 原子呈现+4价. 计算表明Ti 3+—O 键长大于Ti 4+—O 键长, 由于外延生长限制xy 面内的晶格常数, 这导致掺杂层的TiO 6八面体受到了xy 面内的压缩应力, 使得原来简并的d xy , d yz , d xz 三个轨道劈裂, 最终d xy 轨道略高于d xz 和d yz 轨道, 因而电子优先占据d xz 和d yz 轨道. 如图3(b)所示, 在掺杂SrO 层一侧掺杂电子分布在Ti d xz 轨道, 而在另一侧则分布在d yz 轨道. 通过对多种自旋构型的计算比较, 结果表明图3(b)所示的反铁磁自旋序具备更低的能量. 图3(c)展示了局域态所在的Ti 3+与近邻的6个O 原子的键长, 沿着y , z 方向键长明显大于x 方向, 这与电子占据Ti d xz 和d yz 轨道相吻合. 计算结果还表明Ti 3+与近邻的O 原子的键长也明显大于Ti 4+与近邻的O 原子的键长.从体掺杂的角度看, [SLTO]1|[STO]5结构中La 离子平均体密度为4.17%, 而[SLTO]1|[STO]1结构中La 离子平均体密度为12.5%, 似乎La 离子的体密度与上述金属绝缘体转变有关. 而Tokura等[27]和Okuda 等[28]的实验结果表明在STO 内La 离子体掺杂密度在1.5%—92%区间内体系都呈现金属态. 由此可知, 在STO 中均匀掺杂4.17%的La 会导致金属态. 而d 掺杂的[SLTO]1|[STO]5超晶格结构中La 离子平均体密度同为4.17%, 却呈现绝缘体性, 这意味着掺杂维度变化是导致上述金属绝缘体转变的决定因素. 图4展示了三维掺杂和二维掺杂情况下杂质离子层在空间中产生的电势分布示意图. 在[SLTO]1|[STO]1掺杂情况下,如图4(a), 相邻的杂质离子层较近, 其吸引势相互D O S(a1)-4-20Energy/eV24 /e V(a2)3210-1/e V(c2)3210-1D O S(c1)-4-20Energy/eV24 /e V(d2)3210-1/e V(b2)3210-1D O S(b1)-4-20Energy/eV24D O S(d1)-4-20Energy/eV24图 2 分自旋总态密度图和能带图 (a1), (a2) U = 2 eV, [SLTO]1|[STO]1; (b1), (b2) U = 2 eV, [SLTO]1|[STO]5; (c1), (c2) U =3.7 eV, [SLTO]1|[STO]1; (d1), (d2) U = 3.7 eV, [SLTO]1|[STO]5, 红色箭头所指为带隙内局域态. 图中红线为费米能级, 价带顶部设为能量零点. 态密度图中水平线上部为上自旋态密度, 下部为下自旋态密度Fig. 2. Spin-polarized total densities of states and band structures: (a1), (a2) U = 2 eV, [SLTO]1|[STO]1; (b1), (b2) U = 2 eV,[SLTO]1|[STO]5; (c1), (c2) U = 3.7 eV, [SLTO]1|[STO]1; (d1), (d2) U = 3.7 eV, [SLTO]1|[STO]5, the in-gap localized states are pointed out by the red arrow. The red lines are Fermi level, the top of valence band is set to be zero.dddd(b)(c)Ti 3+Ti 3+2.012.082.022.07Ti 4+Ti 3+2.062.07Ti 4+-11/e Vdfd 23图 3 (a) [SLTO]1|[STO]5能带结构图, 其中带隙内局域态为Ti d xz 和d yz 轨道态. 水平红色虚线为费米能级; (b) [SLTO]1|[STO]5带隙内局域态电荷空间分布, 局域态为Ti d xz 和d yz 轨道态, 上下箭头代表自旋方向; (c)掺杂层局部结构和Ti 3+O 6八面体键长, 沿着y 和z 方向Ti 3+—O 键较长Fig. 3. (a) Band structure of [SLTO]1|[STO]5, in which the in-gap states mainly consist of Ti d xz and d yz orbitals; (b) charge distri-bution of the in-gap states, the charge is mainly localized at Ti d xz and d yz orbitals. The arrows represent spin directions; (c) local structure of the doped layer and bond lengths of Ti 3+—O bonds of the Ti 3+O 6 octehedron.重叠较大, 最终在空间产生较为平缓的势. 而在[SLTO]1|[STO]5掺杂情况下, 如图4(b)所示, 相邻的杂质离子层较远, 其吸引势重叠小, 最终在掺杂层形成势阱, 该势阱束缚了电子在垂直掺杂面方向的运动, 结果电子只能在掺杂层内运动. 通常,电子系统的能量取决于电子在邻近格点间跳跃的动能和电子间排斥势能的总和, 关联性强弱大致取决于电子间排斥势能与电子动能的比值, 比值越大则关联性越强. 相比三维掺杂, 二维掺杂情况下电子在杂质离子层的势阱中运动, 在垂直方向运动受限制, 允许电子跳跃的近邻格点变少, 总动能变小,电子运动关联性变强. 二维体系情况下, 若体系呈现金属态, 即电子可在近邻格点巡游, 则动能较低,但存在两个电子同时占据同一个Ti 原子3d 轨道的几率, 由于Ti 3d 轨道上存在较大的在位库伦排斥能, 这会导致较大的电子间排斥势能, 体系的总能量可能因此更高. 若体系呈现绝缘态, 带隙内局域态电子不能在近邻格点巡游, 则动能较大, 但避免了两个电子同时占据同一个Ti 原子3d 轨道引起的较大的在位库伦排斥能, 这降低了电子间排斥势能, 体系的总能量可能因此更低. 这意味着在同样的在位库伦排斥能情况下, 相比三维电子体系,二维电子体系具有更小的动能, 即更强的关联性,更容易变为绝缘态. 上述计算中得到的SrTiO 3中层状25% La 掺杂导致的金属-绝缘体转变正是电子维度降低导致关联性增强的一个实例.此外, 二维电子的密度也影响着体系关联性.从平均场的角度看, 二维电子体系的电子间排斥势能正比于n 1/2(n 为二维电子密度), 动能正比与n ,则电子间排斥势能与动能比值约为n –1/2[29]. 这意味随着二维掺杂浓度的提高, 关联性会变弱, 体系有可能从绝缘态变为金属态. 实验研究[16]和本文的计算都验证了这一点, 图5所示的态密度和能带结构表明当二维La 掺杂的掺杂浓度为50%时上述超晶格结构呈现金属态.D O S(a)-4-20Energy/eV24 /e V(b)3210-1图 5 50% La 掺杂的[SLTO]1|[STO]5总态密度图(a)和能带结构图(b), 红线为费米能级Fig. 5. Total density of states (a) and band structure (b) of [SLTO]1|[STO]5 with 50% La doping in the doping layer.4 结 论本文利用第一性原理计算研究了d 掺杂的La:SrTiO 3中掺杂维度和浓度变化引起的金属绝缘体转变. 在La 掺杂浓度为25%情况下, 随着掺杂层间隔增加, 即掺杂维度从三维过渡到二维, 体系从金属态过渡到绝缘体态. 二维掺杂在SrTiO 3带隙内产生了局域态, 并且局域态呈现出一定的电荷序、反铁磁自旋序和轨道序. 分析表明, 局域态的电子是由二维体系情况下关联性增强引起的. 此外, 二维掺杂的电子密度也影响着体系的状态, 在二维La 掺杂的结构中掺杂浓度为50%时, 体系又呈现金属态. 本文的研究结果加深了对于过渡金属氧化物中电子关联性与其维度和浓度关系的认识,有助于利用维度和浓度调控过渡金属氧化物电子器件的性能.S r 0.75L a 0.25OT i O 2S rO(a)S r 0.75L a 0.25OT i O 2S rO(b)图 4 掺杂离子层的电势V 和掺杂电荷r 分布示意图 (a) [SLTO]1|[STO]1, 虚线代表单个掺杂层阳离子产生的吸引势, 实线代表相邻掺杂层阳离子吸引势叠加后总的吸引; (b) [SLTO]1|[STO]5Fig. 4. Diagrams of electric potential V and charge distribution: (a) [SLTO]1|[STO]1, dashed lines present the potential produced by a single impurity layer, the solid lines present the total potential of all impurity layers; (b) [SLTO]1|[STO]5.参考文献I mada M, Fujimori A, Tokura Y 1998 Rev. Mod. Phys. 701039[1]D agottoE 2005 Science 309 257[2]T okura Y, Nagaosa N 2000 Science 288 462[3]A hn C H, Triscone J M, Mannhart J 2003 Nature 424 1015[4]M annhart J, Schlom D G 2010 Science 327 1607[5]C amjayi A, Haule K, Dobrosavljević V, Kotliar G 2008 Nat.Phys. 4 932[6]B oris A V, Matiks Y, Benckiser E, et al. 2011 Science 332 937[7]O htomo A, Muller D A, Grazul J L, et al. 2002 Nature 419 378[8]O htomo A, Hwang H Y 2004 Nature 427 423[9]R eyren N, Thiel S, Caviglia A D, et al. 2007 Science 317 1196[10]C aviglia A D, Gariglio S, Reyren N, et al. 2008 Nature 456624[11]L i L, Richter C, Mannhart J, Ashoori R C 2011 Nat. Phys. 7 762[12]B rinkman A, Huijben M, Van Zalk M, et al. 2007 Nat. Mater.6 493[13]A riando, Wang X, Baskaran G, et al. 2011 Nat. Commun. 2188[14]W ang Z, Okude M, Saito M, et al. 2010 Nat. Commum. 1 106 [15]C hoi W S, Lee S, Cooper V R, et al. 2012 Nano Lett. 12 4590[16]K ornblum L 2019 Adv. Mater. Interfaces 6 1900480[17]N iu W, Wang X F, Xu Y B, Zhang R 2021 ACS Appl.Electron. Mater. 3 128[18]M ardegan J R L, Christensen D V, Chen Y Z, et al. 2019 Phys. Rev. B 99 134423[19]S ong Q, Yu T L, Lou X, et al. 2019 Nat. Commum. 10 758[20]C hen J K, Mao W, Gao L, et al. 2019 Adv. Mater. 321905060[21]K resse G, Furthmüller J 1996 Phys. Rev. B 54 11169[22]P erdew J P, Burke K, Ernzerhof M 1996 Phys. Rev. Lett. 77 3865[23]B löchl P E 1994 Phys. Rev. B 50 17953[24]K resse G, Joubert D 1999 Phys. Rev. B 59 1758[25]D udarev S L, Botton G A, Savrasov S Y, et al. 1998 Phys.Rev. B 57 1505[26]T okura Y, Taguchi Y, Okada Y, Fujishima Y, Iye Y 1993 Phys. Rev. Lett. 70 2126[27]O kuda T, Nakanishi K, Miyasaka S, Tokura Y 2001 Phys.Rev. B 63 113104[28]B ruus H, Flensberg K 2004 Many-body Quantum Theory inCondensed Matter Physics - An Introduction (New York: Oxford University Press) p41[29]Tuning metal-insulator transition in d-doped La:SrTiO3 superlattice by varying doping dimensionality andconcentration*Li Yun 1)† Lu Wen -Jian 2)1) (School of Physics and Electronic Engineering, Hanshan Normal University, Chaozhou 521041, China)2) (Institute of Solid State Physics, HFIPS, Chinese Academy of Sciences, Hefei 230031, China)( Received 30 April 2021; revised manuscript received 3 July 2021 )AbstractElectronic properties in d-doped La:SrTiO3 superlattices varying with the doping dimensionality and concentration are systematically studied through using first-principles calculation. The superlattices consist of periodically repeated La-doped single SrTiO3 layers in SrTiO3 film, and the doping dimensionality can be tuned by varying the space of the neighboring doped layers. At 25% doping concentration, the spacing between SrTiO3 layers increases from 1 unit-cell layer to 5 unit-cell layers, i.e. the doping dimensionality changes three dimensions to two dimensions, the superlattice charater changes from metallic character into insulating character, and the charge sequence, spin sequence and orbital sequence are present in a localized state. This metal-insulator transition is ascribed to the stronger correlation effect in the two-dimensional electron system. With the two-dimensional doping concentration increasing to 50%, the correlation effect becomes weak and the system becomes metallic.Keywords: two dimensional doping, strong correlation effect, metal-insulator transitionPACS: 71.30.+h, 73.20.–r, 71.27.+a DOI: 10.7498/aps.70.20210830* Project supported by the Scientific Research Staring Foundation for the Returned Overseas Chinese Scholars, Ministry of Education of China (Grant No. [2015]-1098).† Corresponding author. E-mail: liyunphy@。
a r X i v :0705.1627v 1 [q u a n t -p h ] 11 M a y 2007Nonlinear Coherent Destruction of TunnelingXiaobing Luo,Qiongtao Xie,and Biao Wu ∗Institute of Physics,Chinese Academy of Sciences,Beijing 100080,ChinaWe study theoretically two coupled periodically-curved optical waveguides with Kerr nonlinearity.We find that the tunneling between the waveguides can be suppressed in a wide range of parameters due to nonlinearity.Such suppression of tunneling is different from the coherent destruction of tun-neling in a linear medium,which occurs only at the isolated degeneracy point of the quasienergies.We call this novel suppression nonlinear coherent destruction of tunneling.This nonlinear phe-nomenon can be observed readily with current experimental capability;it may also be observable in a different physical system,Bose-Einstein condensate.PACS numbers:42.65.Wi,42.82.Et,03.75.Lm,33.80.BePeriodic driving force is an important and effective tool for coherently controlling quantum tunneling.This has been well demonstrated with a paradigmatic model,a free particle in a double-well potential and driven by a periodic external field[1].With appropriately tuned pa-rameters,the periodic driving force is able not only to enhance tunneling[2]-[4]but also to completely suppress it[5]-[8].The latter is rather surprising and was discov-ered first by Grossmann et al [5].It is now known as coherent destruction of tunneling (CDT)[5].When it oc-curs,a localized wave packet prepared in one well re-mains in the same well and does not tunnel to the other well.In a periodically driven system,there are Floquet states and associated quasienergies[9].The CDT is found to occur only at the isolated degeneracy point of the quasienergies[5,6].Recently,this quantum phenomenon of CDT was ob-served experimentally with two coupled periodically-curved waveguides[10](see Fig.1).In this classical optical system,the Maxwellian wave mimics the quantum wave while the periodic driving force is achieved by bending the waveguides periodically.Such a waveguide system is an ideal laboratory system for demonstrating the co-herent control of quantum tunneling by periodic driving force.For example,tunneling enhancement has recently also been reported with two optical waveguides[11].In this Letter we consider a similar coupled waveguide system but with Kerr nonlinearity.With a well-known two-mode approximation,the system can be described by a two-mode nonlinear model with an external peri-odic driving force.This driving is characterized by two parameters,its frequency w (the inverse of the period of the curved waveguide)and its strength S (the curving magnitude of the waveguides)of the driving force.By numerically solving this two-mode nonlinear model,we find that the suppression of tunneling between the two coupled waveguides happens for a wide range of ratio S/w .This is in stark contrast to the CDT in curved lin-ear waveguides that occurs at an isolated point of S/w ,where the quasienergies of the system are degenerate.This extension of tunneling suppression region is caused by nonlinearity.Therefore,we call it nonlinear coherentdestruction of tunneling (NCDT).We find that the range of ratio S/w for NCDT increases steeply with nonlinear strength.The Floquet states and the quasienergies of this nonlinear model are also studied.We discover that there can be more than two Floquet states and quasiener-gies in a certain range of ratio S/w .These additional Floquet states form a triangle in the quasienergy levels.Our study reveals that these additional Floquet states are closely related to the NCDT.The current experimental capability with nonlinear waveguides is examined.We find that the observation of NCDT is well within the current experimental abil-ity.Note that the nonlinear two-mode model that we de-rived for the waveguides can also be used to describe the dynamics of a Bose-Einstein condensate in a double-well potential under a periodic modulation[12].This indicates that NCDT may also be observable with Bose-Einstein condensates.FIG.1:Schematic drawing (not to scale)of two periodically curved optical waveguides placed parallel to each other.In a weakly guiding dielectric structure,the effective two-dimensional wave equation for light propagation in nonlinear directional waveguides reads[13]i λ∂z=−λ2∂x 2+V [x −x 0(z )]ψ−|ψ|2ψ.(1)where λis the free space wavelength of the light,x 0(z )=A cos(2πz/Λ),and V (x )≡[n 2s −n 2(x )]/(2n s )≃n s −n (x ),where n (x )and n s are,respectively,the effective refractive index profile of the waveguides and the sub-strate refractive index.For the coupled waveguides as in2Fig.1,n (x )thus V (x )have a double-well structure.The scalar electric field is related to ψthrough E (x,z,t )=(1/2)(|n 2|n s ǫ0c 0/2)−1/2[ψ(x,z )exp(−iωt +ikn s z )+c.c.],where n 2is the nonlinear refractive index of the medium,k =2π/λ,ω=kc 0,and c 0and ǫ0are the speed of light and the dielectric constant in vacuum,respectively.The field normalization is taken such that |ψ|2/|n 2|gives the light intensity I (in W/m 2).By means of a Kramers-Henneberger transformation[14]x ′=x −x 0(z ),z ′=z ,and φ(x ′,z ′)=ψ(x ′,z ′)exp[−i (2n s π/λ)˙x 0(z ′)x ′−i (n s π/λ) z′0dξ˙x 20(ξ)](the dot indicates the derivativewith respect to z ′),Eq.(1)is then transformed to i λ∂z ′=−λ2∂x ′2+V (x ′)φ−|φ|2φ+x ′F (z ′)φ≡H 0φ−|φ|2φ+x ′F (z ′)φ.(2)where F (z ′)=n s ¨x 0(z ′)=(4π2An s /Λ2)cos(2πz ′/Λ)is the force induced by waveguide bending.It is clear that if we view z (or z ′)as time t ,the above equations can be regarded as describing the system of a nonlinear quan-tum wave in a double-well potential and under a periodic modulation.We assume that the light in each waveguide of the cou-pler is single moded and neglect excitation of radiation modes.With a standard two-mode approximation[15,16,17],we write φ(x ′,z ′)=e −2iπD exp[−(x ′±a/2)2/2b 2],wherea is the distance between the two waveguides,b is the half-width of each waveguide,and D is related to the input power of the system P (0)as D =n 2P (0)/(√2c 2−S 2c 1+S2πn 2P (0)/(λb )is an effective nonlinearcoefficient.When S =0,Eqs.(4),(5)will be reduced to the well-known Jensen equation[17].Note that P (0)has the unit of W/m is because the waveguide is two dimensional in our theoretical model.In experiments,P (0)has the unit of W and the waveguides are three dimensional.As a result,to relate our nonlinear pa-rameter to the real experimental parameters,we chooseFIG.2:The intensity of light in the initially populated waveg-uide for the case of χ=0(dashed lines)and χ/v =0.4(solid lines)with (a)S/w =1.8,(b)S/w =2.2,(c)S/w =2.4.Dis-tance z ′is in units of 1/v .w/v =10.χ=2πn 2P (0)/(λσeff),where σeffis the effective cross-section of the waveguide,according to Ref.[18].To investigate tunneling effect,we solve the two non-linear equations (4)and (5)numerically with the light initially localized in one of the two waveguides.With the numerical solution,we compute the intensity of the lightstaying in the initial well with P ′(z ′)=|c ∗1(0)c 1(z ′)+c ∗2(0)c 2(z ′)|2.Three sets of our results are shown in Fig.2(a,b,c).In the first set for S/w =1.8,we see that P ′(z ′)oscillates between zero and one for both linear case χ=0and nonlinear case χ/v =0.4,demonstrating no suppression of tunneling.In the second set for S/w =2.2,we see a different scenario,the oscillation of P ′(z ′)is lim-ited between ∼0.8and one for the nonlinear case,showing suppression of tunneling,while there is no suppression for the linear case.In the third set for S/w =2.4,sup-pression of tunneling is seen for both linear and nonlinear cases.Such suppression of tunneling for the linear case is known as coherent destruction of tunneling[5].These nu-merical results demonstrate that nonlinearity can extend the parameter range of the suppression of tunneling.We call this new phenomenon nonlinear coherent destruction of tunneling (NCDT).The extension of tunneling suppression regime of ra-tio S/w by nonlinearity is more clearly demonstrated in Fig.3(a).In this figure,we have used localization,which is defined as the minimum value of P ′(z ′),to measure the suppress of tunneling.When there is large suppres-sion of tunneling,localization is close to one;when there is no suppression,localization is zero.As clearly seen in Fig.3(a),the peak of localization (solid line)for χ/v =0.4is much wider than the peak for χ/v =0.0(dashed line).In Fig.3(b),we see the width of localization ∆Γincreases almost linearly with nonlinearity χ(solid line).Note that,analytically,CDT occurs only at isolated points.That it3FIG.3:(a)Localization as a function of S/w .The solid line isfor the nonlinear case χ/v =0.4and the dashed line is for the linear case χ=0.w/v =10.(b)The width ∆Γof the peak in (a)as a function of nonlinearity strength χ/v (solid line).Thedashed line is for the width of the quasienergy triangle in Fig.4.has a narrow range in Fig.3(a)is because the evolution time is finite in numerical simulation.As is well known,the CDT is connected to the degen-eracy point of quasienergies in the system[5].Although our system is nonlinear,one can similarly define its Flo-quet state and quasienergy.That is,Eqs.(4,5)have so-lutions in the form of {c 1,c 2}=e −iεz ′{˜c 1(z ′),˜c 2(z ′)},where both ˜c 1and ˜c 2are periodic with period of Λ.These Floquet states and corresponding quasienergies εcan be found numerically.We first expand the periodic functions ˜c 1,2in terms of Fourier series with a cutoff.After plug-ging them into Eqs.(4,5),we obtain a set of nonlinear equations for the Fourier coefficients.By solving these equations numerically,we obtain the Floquet states and corresponding quasienergies ε.The results are plotted in Fig.4,where we witness a striking difference between the linear and nonlinear cases.As seen in Fig.4(a),for the linear case,there are two Floquet states for a given value of S/w and there is only one isolated degeneracy point.For the nonlinear case,we notice that there are four Flo-quet states and three quasienergies in a certain range of S/w with two of the Floquet states degenerate.The three quasienergies form a triangle in the quasienergy levels as seen in Fig.4(b,c).Our numerical computation shows that the width of the quasienergy triangle increases with nonlinearity χas shown in Fig.3(dashed line).As this increasing trend is similar to the localization width ∆Γ,this offers us the first glimpse of link between NCDT and the quasienergies.Since the right corner of the triangle can be open,we define the width of the quasienergy tri-angle as the horizontal distance between the left corner and the upper corner.A firm link between the NCDT and the triangle struc-ture in the quasienergies can be established by looking into the Floquet states.We focus on the Floquet statesFIG.4:Quasienergies at (a)χ=0;(b)χ/v =0.4;(c)χ/v =0.8.Solid lines are for numerical results obtained with Eqs.(4,5)and circles for the approximation results for high frequencies with Eqs.(7,8).w/v =3.that correspond to the lowest quasienergies in Fig.4.To measure how the Floquet state is localized in one of thetwo waveguides,we define |c 1|2 =( Λ0dz ′|c 1|2)/Λfor a given Floquet state {c 1,c 2}.We have plotted this value for the lowest Floquet states in Fig.5.In this figure,we see clearly that only the Floquet states on the quasienergy triangle are localized.This thus demon-strates a clear link between the quasi-energy triangle and the NCDT.That there are two lines in Fig.5reflects the fact that there is a two-fold degeneracy for the lowest quasienergies on the triangle.FIG.5:Intensity in the first well for every Floquet state in the lowest quasienergy level at χ/v =0.4,w/v =3.The triangular structure in the quasienergy is very similar to the energy loop discovered within the con-text of nonlinear Landau-Zener tunneling[21].In fact,they are mathematically related.For high frequencies,w ≫max {v,χ},which is usually the case for current ex-periments with optical waveguides,we take advantage of the transformationc 1=c ′1exp[iS sin(wz ′)/2w ],c 2=c ′2exp[−iS sin(wz ′)/2w ].(6)4After averaging out the high frequency terms[12],wefind a non-driving nonlinear model,i˙c′1=v2J0(S/w)c′1−χ|c′2|2c′2,(8)where J0is the zeroth-order Bessel function.It is clear from the transformation in Eq.(6)that the eigenstates of the above time-independent nonlinear equations cor-respond to the Floquet states of Eqs.(4,5).We have com-puted the eigenstates of Eqs.(7,8)and the corresponding eigenenergies,which are plotted as circles in Fig.4.The consistency with the previous results is obvious.As is known in Ref.[21],the above nonlinear model admits ad-ditional eigenstates whenχ>J0(S/w)v.Therefore,this can be regarded as the condition for the extra Floquet states to appear for the driving nonlinear model Eqs.(4,5) at high frequencies.So far,we have focused on self-focusing materials.Our approach and results will be very similar if one consid-ers instead self-defocusing materials,for which the sign before the nonlinear term in Eq.(1)should be plus.Non-linear coherent destruction of tunneling still occurs and the triangular structure also appears in the quasienergy levels but its direction is reversed as compared to the self-focusing case.At present the nonlinear waveguides are readily avail-able in labs[18,19,20].We take the experimental parameters in Ref.[19]to estimate our theoretical val-ues in Eqs.(4,5).The wavelength of the laser light is λ=1.55µm,the effective cross-sectional area of the waveguide isσeff=12µm2,the nonlinear index n2= 1.2×10−13cm2/W,and the shortest length for the light transfer from one waveguide to the other waveguide in the weak nonlinearity limit is L c≈2cm.With the power input in the waveguides P(0)∼100W,we haveχπλσeff≈2.(9) This shows that strong nonlinear waveguides are avail-able at optical labs and nonlinear coherent destruction of tunneling can be visualized in an optical experiment similar to the one in Ref.[10].We also want to mention briefly that NCDT may be applied to improve optical switching devices[19,20].The details will be discussed elsewhere.In conclusion,we have studied the light propagation in a nonlinear periodically-curved waveguide directional coupler.We have found a new type of suppression of tunneling in this system,which is induced by nonlinear-ity and has no linear counterpart.We call it nonlinear coherent destruction of tunneling(NCDT)in analogy to a similar but different phenomenon in linear driving sys-tems,coherent destruction of tunneling.The NCDT oc-curs for an extended range of ratio S/w,where S is the strength of the driving and w is its frequency.We have also found that the NCDT is closely related to a trian-gular structure appeared in the quasienergy levels of the nonlinear system.We have also pointed out that obser-vation of the novel nonlinear phenomenon is well within the capacity of current experiments.This work is supported by NSF of China(10504040), the973project of China(2005CB724500,2006CB921400), and the“BaiRen”program of Chinese Academy of Sci-ences.∗Electronic address:bwu@[1]M.Grifoni,and P.H¨a nggi,Phys.Rep.304,229(1998).[2]W.A.Lin and L.E.Ballentine,Phys.Rev.Lett.65,2927(1990).[3]A.Peres,Phys.Rev.Lett.67,158(1991).[4]I.Vorobeichik and N.Moiseyev,Phys.Rev.A,59,2511(1999).[5]F.Grossmann,T.Dittrich,P.Jung,and P.Hanggi,Phys.Rev.Lett.67,516(1991);Z.Phys.B84,315(1991). [6]F.Grossmann and P.Hanggi,Europhys.Lett.18,571(1992).[7]R.Bavli and H.Metiu,Phys.Rev.Lett.69,1986(1992).[8]M.Steinberg and U.Peskin,J.Appl.Phys.85,270(1999).[9]J.H.Shirley,Phys.Rev.138,B979(1965).[10]G.Della Valle,M.Ornigotti, E.Cianci,V.Fogli-etti,porta,and S.Longhi.e-print arXiv:quant-ph/0701121.[11]I.Vorobeichik,E.Narevicius,G.Rosenblum,M.Oren-stein,and N.Moiseyev,Phys.Rev.Lett.90,176806 (2003).[12]Guan-Fang Wang,Li-Bin Fu and Jie Liu,Phys.Rev.A73,013619(2006).[13]R.W.Micallef,Y.S.Kivshar,J.D.Love,D.Burak,andR.Binder,Opt.Quantum Electron.30,751(1998).[14]W.C.Henneberger,Phys.Rev.Lett.21,838(1968).[15]S.Longhi,Phys.Rev.A71,065801(2005).[16]R.Khomeriki,J.Leon,and S.Ruffo,Phys.Rev.Lett.97,143902(2006).[17]S.M.Jensen,IEEE J.Quantum Electron.QE-18,1580(1982).[18]H.S.Eisenberg,Y.Silberberg,R.Morandotti, A.R.Boyd,and J.S.Aitchison,Phys.Rev.Lett.81,3383 (1998).[19]K.Al-hemyari,A.Villeneuve,J.U.Kang,J.S.Aitchison,C.N.Ironside,G.I.Stegeman,Appl.Phys.Lett.63,3562(1993).[20]S.R.Friberg,Y.Silberberg,M.K.Oliver,M.J.Andrejco,M.A.Saifi,P.W.Smith,Appl.Phys.Lett.51,135(1987).[21]B.Wu and Q.Niu,Phys.Rev.A61,023402(2000).。
道明光学超导概念
道明光学超导（Demonmetiou optic superconductivity）是一种
在光学材料中观察到的现象，它类似于超导体中的电子对的库伦耦合。
它是由斯蒂尔和鲁卡以及道赛父子的研究团队发现的。
道明光学超导的基本概念是在介质中通过光的相互作用形成光子对，这些光子对在材料中传递，并表现出类似超导体中电子对的特性。
这些光子对可以通过相干的相互作用来传输能量和信息，而且它们之间有一种强相互作用。
这种相互作用类似于超导体中的库伦相互作用，它使得光子对能够在材料中几乎无阻碍地传递。
与传统的光学材料不同，道明光学超导体具有较低的折射率，这是由于光子对的强耦合导致的。
这种强耦合可以在一定的温度和光强度下产生，并具有与超导体相似的零电阻和非局域性等特性。
道明光学超导具有许多潜在的应用领域，包括高速光通信、量子计算、光量子存储等。
然而，目前对于道明光学超导的研究还处于起步阶段，需要进一步的实验和理论研究来深入了解和应用这种现象。
第４０卷㊀第７期２０１９年７月发㊀光㊀学㊀报ＣＨＩＮＥＳＥＪＯＵＲＮＡＬＯＦＬＵＭＩＮＥＳＣＥＮＣＥＶｏｌ ４０Ｎｏ ７Ｊｕｌｙꎬ２０１９文章编号:１０００￣７０３２(２０１９)０７￣０９１５￣０７界面处理对ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓ器件动态特性的影响韩㊀军１ꎬ赵佳豪１ꎬ赵㊀杰１ꎬ２ꎬ邢艳辉１∗ꎬ曹㊀旭１ꎬ付㊀凯２ꎬ宋㊀亮２ꎬ邓旭光２ꎬ张宝顺２(１.北京工业大学信息学部光电子技术省部共建教育部重点实验室ꎬ北京㊀１００１２４ꎻ２.中国科学院苏州纳米技术与纳米仿生研究所纳米器件与应用重点实验室ꎬ江苏苏州㊀２１５１２３)摘要:研究不同界面处理对ＡｌＧａＮ/ＧａＮ金属￣绝缘层￣半导体(ＭＩＳ)结构的高电子迁移率晶体管(ＨＥＭＴ)器件性能的影响ꎮ采用Ｎ２和ＮＨ３等离子体对器件界面预处理ꎬ实验结果表明ꎬＮ２等离子体预处理能够减小器件的电流崩塌ꎬ通过对Ｎ２等离子体预处理的时间优化ꎬ发现预处理时间１０ｍｉｎ能够较好地提高器件的动态特性ꎬ３０ｍｉｎ时动态性能下降ꎮ进一步引入ＡｌＮ作为栅介质插入层并经过高温热退火后能够有效提高器件的动态性能ꎬ将器件的阈值回滞从４１１ｍＶ减小至１１１ｍＶꎬ动态测试表明ꎬ在９００Ｖ关态应力下ꎬ器件的电流崩塌因子从４２.０４减小至４.７６ꎮ关㊀键㊀词:电流崩塌ꎻＡｌＮ栅介质插入层ꎻ界面处理ꎻＡｌＧａＮ/ＧａＮ高电子迁移率晶体管中图分类号:ＴＮ３８６.２㊀㊀㊀文献标识码:Ａ㊀㊀㊀ＤＯＩ:１０.３７８８/ｆｇｘｂ２０１９４００７.０９１５ＩｍｐａｃｔｏｆＩｎｔｅｒｆａｃｅＴｒｅａｔｍｅｎｔｏｎＤｙｎａｍｉｃＣｈａｒａｃｔｅｒｉｓｔｉｃｏｆＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓＨＡＮＪｕｎ１ꎬＺＨＡＯＪｉａ￣ｈａｏ１ꎬＺＨＡＯＪｉｅ１ꎬ２ꎬＸＩＮＧＹａｎ￣ｈｕｉ１∗ꎬＣＡＯＸｕ１ꎬＦＵＫａｉ２ꎬＳＯＮＧＬｉａｎｇ２ꎬＤＥＮＧＸｕ￣ｇｕａｎｇ２ꎬＺＨＡＮＧＢａｏ￣ｓｈｕｎ２(１.ＫｅｙＬａｂｏｒａｔｏｒｙｏｆＯｐｔｏ￣ｅｌｅｃｔｒｏｎｉｃｓＴｅｃｈｎｏｌｏｇｙꎬＭｉｎｉｓｔｒｙｏｆＥｄｕｃａｔｉｏｎꎬＢｅｉｊｉｎｇＵｎｉｖｅｒｓｉｔｙｏｆＴｅｃｈｎｏｌｏｇｙꎬＢｅｉｊｉｎｇ１００１２４ꎬＣｈｉｎａꎻ２.ＫｅｙＬａｂｏｒａｔｏｒｙｏｆＮａｎｏＤｅｖｉｃｅｓａｎｄＡｐｐｌｉｃａｔｉｏｎｓꎬＳｕｚｈｏｕＩｎｓｔｉｔｕｔｅｏｆＮａｎｏ￣ｔｅｃｈａｎｄＮａｎｏ￣ｂｉｏｎｉｃｓꎬＣｈｉｎｅｓｅＡｃａｄｅｍｙｏｆＳｃｉｅｎｃｅｓꎬＳｕｚｈｏｕ２１５１２３ꎬＣｈｉｎａ)∗ＣｏｒｒｅｓｐｏｎｄｉｎｇＡｕｔｈｏｒꎬＥ￣ｍａｉｌ:ｘｉｎｇｙａｎｈｕｉ＠ｂｊｕｔ.ｅｄｕ.ｃｎＡｂｓｔｒａｃｔ:ＴｈｅｅｆｆｅｃｔｓｏｆｄｉｆｆｅｒｅｎｔｋｉｎｄｓｏｆｉｎｔｅｒｆａｃｅｔｒｅａｔｍｅｎｔｏｎｔｈｅｃｈａｒａｃｔｅｒｉｓｔｉｃｏｆＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓｗｅｒｅｓｔｕｄｉｅｄｉｎｔｈｉｓｐａｐｅｒ.Ｎ２ａｎｄＮＨ３ｐｌａｓｍａｐｒｅｔｒｅａｔｍｅｎｔｗｅｒｅｕｓｅｄｔｏｉｍｐｒｏｖｅｔｈｅｉｎｔｅｒｆａｃｅｑｕａｌｉｔｙ.ＴｈｅｒｅｓｕｌｔｓｓｈｏｗｔｈａｔＮ２ｐｌａｓｍａｐｒｅｔｒｅａｔｍｅｎｔｃｏｕｌｄｒｅｄｕｃｅｔｈｅｃｕｒｒｅｎｔｃｏｌｌａｐｓｅｏｆｄｅｖｉｃｅｓ.ＢｙｏｐｔｉｍｉｚｉｎｇｔｈｅｔｉｍｅｏｆＮ２ｐｌａｓｍａｐｒｅｔｒｅａｔｍｅｎｔꎬｉｔｗａｓｆｏｕｎｄｔｈａｔｔｈｅｄｙｎａｍｉｃｃｈａｒａｃｔｅｒｉｓｔｉｃｏｆｄｅｖｉｃｅｓｗｉｔｈ１０ｍｉｎｔｈｅｐｒｅｔｒｅａｔｍｅｎｔｗａｓｉｍｐｒｏｖｅｄꎬｗｈｉｌｅｔｈａｔｏｆ３０ｍｉｎｗａｓｄｅｇｒａｄｅｄ.Ａｓａｇａｔｅｄｉｅｌｅｃｔｒｉｃｉｎ￣ｔｅｒｃａｌａｔｉｏｎｌａｙｅｒꎬｔｈｅａｎｎｅａｌｅｄＡｌＮｉｎｔｅｒｌａｙｅｒｃａｎｅｆｆｅｃｔｉｖｅｌｙｉｍｐｒｏｖｅｔｈｅｄｙｎａｍｉｃｃｈａｒａｃｔｅｒｉｓｔｉｃｏｆｔｈｅｄｅｖｉｃｅ.ＴｈｅＶｔｈｈｙｓｔｅｒｅｓｉｓｗａｓｄｅｃｒｅａｓｅｄｆｒｏｍ４１１ｍＶｔｏ１１１ｍＶꎬａｎｄｔｈｅｄｅｖｉｃｅｃｕｒｒｅｎｔｃｏｌｌａｐｓｅｆａｃｔｏｒｗａｓｒｅｄｕｃｅｄｆｒｏｍ４２.０４ｔｏ４.７６ａｆｔｅｒｕｎｄｅｒＯＦＦ￣ｓｔａｔｅＶＤｓｔｒｅｓｓｏｆ９００.Ｋｅｙｗｏｒｄｓ:ｃｕｒｒｅｎｔｃｏｌｌａｐｓｅꎻＡｌＮｇａｔｅｄｉｅｌｅｃｔｒｉｃｉｎｓｅｒｔｉｏｎｌａｙｅｒꎻｉｎｔｅｒｆａｃｅｔｒｅａｔｍｅｎｔꎻＡｌＧａＮ/ＧａＮｈｉｇｈｅｌｅｃｔｒｏｎｍｏｂｉｌｉｔｙｔｒａｎｓｉｓｔｏｒｓ㊀㊀收稿日期:２０１８￣０８￣２０ꎻ修订日期:２０１８￣１０￣１７㊀㊀基金项目:国家自然科学基金(６１２０４０１１ꎬ１１２０４００９ꎬ６１５７４０１１)ꎻ北京市自然科学基金(４１４２００５ꎬ４１８２０１４)ꎻ北京市教委科学研究基金(ＰＸＭ２０１８＿０１４２０４＿５０００２０)资助项目ＳｕｐｐｏｒｔｅｄｂｙＮａｔｉｏｎａｌＮａｔｕｒａｌＳｃｉｅｎｃｅＦｏｕｎｄａｔｉｏｎｏｆＣｈｉｎａ(６１２０４０１１ꎬ１１２０４００９ꎬ６１５７４０１１)ꎻＢｅｉｊｉｎｇＮａｔｕｒａｌＳｃｉｅｎｃｅＦｏｕｎｄａ￣ｔｉｏｎ(４１４２００５ꎬ４１８２０１４)ꎻＢｅｉｊｉｎｇＭｕｎｉｃｉｐａｌＥｄｕｃａｔｉｏｎＣｏｍｍｉｓｓｉｏｎＳｃｉｅｎｔｉｆｉｃＲｅｓｅａｒｃｈＦｕｎｄ(ＰＸＭ２０１８＿０１４２０４＿５０００２０)９１６㊀发㊀㊀光㊀㊀学㊀㊀报第４０卷１㊀引㊀㊀言ＧａＮ作为第三代半导体的代表ꎬ具有高禁带宽度㊁高击穿电场㊁高电子迁移率㊁以及耐酸碱等特点ꎮ以ＡｌＧａＮ和ＧａＮ异质结结构制备的高电子迁移率晶体管ꎬ由于极化效应产生的天然的高浓度㊁高迁移率的二维电子气ꎬ在功率开关器件的大功率及高频性能方面有很好的应用前景[１￣４]ꎮＭＩＳ￣ＨＥＭＴ器件可以有效地减小器件的栅极漏电ꎬ提高耐压ꎬ提高栅驱动能力ꎮ但是由于栅介质的引入ꎬ产生新的界面ꎬ界面质量给器件的应用带来新的问题ꎬ影响器件的可靠性和阈值回滞等ꎮＥｌｌｅｒ等[５]详细报道了对于ＧａＮ表面的处理过程ꎬ包括湿法化学处理[６]㊁真空退火处理[７]㊁气体氛围下退火处理[８]及离子束㊁等离子体处理[９￣１０]等ꎮＧａＮ材料表面存在含Ｏ的化合物和Ｎ空位[２ꎬ１１]ꎬ这两种缺陷态成为影响界面质量的主要因素ꎬ目前的报道中ꎬ集中于使用含Ｎ等离子体来处理器件表面[１２￣１４]ꎬ主要作用机理为去除Ｏ杂质和补充Ｎ空位ꎮＨａｓｈｉｚｕｍｅ[１５]在器件钝化作用前使用Ｎ２作为等离子体处理样品表面ꎬ得到了很高质量的钝化结果ꎬ而且界面态浓度下降ꎮＲｏｍｅｒｏ[１６]通过原位含氮气等离子体预处理ꎬ器件的电流崩塌㊁输出功率㊁增益等特性取得了非常好的效果ꎮ在本文研究中ꎬ我们对ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件工艺过程中的界面处理进行优化比较ꎬ实验利用等离子体预处理研究不同气体(Ｎ２和ＮＨ３)及不同预处理时间对器件直流性能和动态特性的影响ꎬ并在该研究基础上ꎬ继续引入ＡｌＮ栅介质插入层进行界面处理ꎬ研究采用ＡｌＮ栅介质插入层进行界面处理对器件动静态特性的影响ꎮ２㊀实㊀㊀验ＡｌＧａＮ/ＧａＮＨＥＭＴ外延材料是通过金属有机物化学气相沉积技术在Ｓｉ(１１１)衬底上生长的ꎬ外延结构依次为成核层㊁ＧａＮ缓冲层和ＡｌＧａＮ势垒层ꎮ器件的制备工艺过程为:(１)界面处理过程ꎻ(２)栅介质钝化层制备ꎬ采用ＬＰＣＶＤ沉积ＳｉＮｘ作为栅介质ꎬ主要考虑其具有良好的稳定性和漏电[７]ꎬ利用ＳｉＨ２Ｃｌ２和ＮＨ３作为Ｓｉ源和Ｎ源ꎬ温度７８０ħꎻ(３)注入隔离ꎬ采用Ｆ离子进行注入隔离ꎻ(４)欧姆接触制备ꎬ利用磁中性环路放电刻蚀ＳｉＮｘ形成窗口ꎬ电子束蒸发沉积Ｔｉ/Ａｌ/Ｎｉ/Ａｕ为２０/１３０/５０/５０ｎｍꎬＮ２氛围下８５０ħ退火３０ｓ形成欧姆接触ꎻ(５)栅电极制备ꎬ利用金属热蒸发沉积Ｎｉ/Ａｕ为５０/１０ｎｍ制备栅电极ꎮ图１(ａ)显示的是ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件基本结构示意图ꎬ器件栅介质层厚度为２０ｎｍꎬ器件栅长为２μｍꎬ栅宽为１００μｍꎬ栅漏距离为１６μｍꎬ栅源距离为４μｍꎮ其中对于界面处理工艺过程ꎬ设计了实验Ⅰ:采用不同预处理气体Ｎ２和ＮＨ３对ＡｌＧａＮ/ＧａＮＨＥＭＴ表面预处理ꎬ预处理时间均为５ｍｉｎꎬ实验分别设置为样品Ａ和样品Ｂꎮ在实验Ｉ基础上设计实验方案Ⅱ:选取Ｎ２作为预处理气体ꎬ研究不同预处理时间对ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件的影响ꎬ设置样品Ｃ㊁Ｄ㊁Ｅ分别预处理的时间为０ꎬ１０ꎬ３０ｍｉｎꎮ上述等离子体预处理温度为３５０ħꎬ压强为２６６Ｐａ(２０００ｍｔｏｒｒ)ꎬＲＦ功率为６０ＷꎬＬＦ功率为５０ＷꎮSiN x2DEGAlGaNSiBufferSGAlN2DEGAlGaNSiN xSiBufferSGDD（a）（b）图１㊀(ａ)实验器件基本结构示意图ꎻ(ｂ)引入插入层后的器件结构示意图ꎮＦｉｇ.１㊀(ａ)Ｓｃｈｅｍａｔｉｃｏｆｄｅｖｉｃｅｓｆｏｒｄｉｆｆｅｒｅｎｔｐｒｅ￣ｔｒｅａｔｍｅｎｔ.(ｂ)Ｓｃｈｅｍａｔｉｃｏｆｄｅｖｉｃｅｓｔｒｕｃｔｕｒｅｆｏｒｓａｍｐｌｅｗｉｔｈｉｎ￣ｓｅｒｔｉｏｎｌａｙｅｒ.为进一步改善ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件性能ꎬ在上述实验的基础上ꎬ设计实验Ⅲ:采取ＰＥＡＬＤ生长的ＡｌＮ作为栅介质插入层ꎬ设置样品Ｆ㊁Ｇ㊁Ｈꎬ引入ＡｌＮ插入层的器件结构示意图为图１(ｂ)ꎮ样品Ｆ作为空白对照组未引入插入层界面处理过程ꎬ样品Ｇ和样品Ｈ利用ＰＥＡＬＤ生长３ｎｍＡｌＮꎬＴＭＡｌ为Ａｌ源ꎬＮ２为Ｎ源ꎬ生长温度３００ħꎮ样品Ｈ在栅介质沉积后于Ｎ２氛围下１０００ħ退火２ｍｉｎꎮ样品栅介质ＬＰＣＶＤ￣ＳｉＮｘ１２ｎｍꎮ器件尺寸分别为:栅长２μｍꎬ栅宽１００μｍꎬ栅漏距离３０μｍꎬ栅源距离３μｍꎮ每组实验均采用安捷伦Ｂ１５０５Ａ进行测试表征ꎮ㊀第７期韩㊀军ꎬ等:界面处理对ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓ器件动态特性的影响９１７㊀３㊀结果与讨论３.１㊀界面预处理气体的影响图２是Ｎ２和ＮＨ３预处理器件的转移输出曲线ꎬ从图２中可以看出不同的预处理气体对器件的直流特性具有明显的影响ꎮＮ２和ＮＨ３等离子体预处理之后器件的峰值跨导分别是６４.６ｍＳ/ｍｍ和７０.７ｍＳ/ｍｍꎬ饱和电流分别为５７９.３ｍＡ/ｍｍ和５５０ｍＡ/ｍｍꎮＮ２等离子体预处理的器件跨导峰值较ＮＨ３等离子体预处理器件低ꎬ但是饱和电流有所增加ꎮ在图２中还看到ꎬ相比于Ｎ２等离子体预处理ꎬＮＨ３等离子体预处理的实验结果中存在饱和电流下降的现象ꎬ这与Ｋｉｍ[１２]报道的一致ꎬ究其原因是在ＮＨ３在较低功率下产生等离子体的同时会产生一个Ｈ＋的钝化效果ꎮ类似的钝化对于器件的ＲＦ性能会有所提升ꎬ但对器件的ＤＣ特性有退化ꎬＨａｓｈｉｚｕｍｅ[１７]和Ｒｏｍｅｒｏ[１６]的研究已经证明了这一点ꎮ为了进一步对比采用Ｎ２和ＮＨ３不同预处理气体对表面态引起的器件性能退化作用ꎬ实验对样品Ａ和样品Ｂ进行了电流崩塌的表征ꎮ图３分别显示了关态下漏极电压600500-20V GS /VI D /(m A ·m m -1)4003002001000N 2plasmaNH 3plasma（a ）-15-10-55406080200G m /(m S ·m m -1)6005002V d /VI D /(m A ·m m -1)4003002001000N 2plasma NH 3plasma（b ）461012143V -3V -5V -7V -9V80V GS -15~3V 图２㊀Ｎ２和ＮＨ３等离子体预处输出曲线理器件转移输出曲线对比ꎮ(ａ)转移曲线ꎻ(ｂ)输出曲线ꎮＦｉｇ.２㊀ＴｔｒａｎｓｆｅｒａｎｄｏｕｔｐｕｔｃｕｒｖｅｓｆｏｒｓａｍｐｌｅＡｗｉｔｈＮ２ｐｌａｓｍａａｎｄｓａｍｐｌｅＢｗｉｔｈＮＨ３ｐｌａｓｍａ.(ａ)Ｔｒａｎｓ￣ｆｅｒｃｕｒｖｅｓ.(ｂ)Ｏｕｔｐｕｔｃｕｒｖｅｓ.10080300V d /VR D y n a m i c /R O N20010100506040200N 2plasma NH 3plasmaOFF 鄄state:V GS =-15VOFF 鄄ON swtiching time:t =200滋s ON 鄄state:V GS =0V V D =1V图３㊀Ｎ２和ＮＨ３等离子体预处理器件电流崩塌对比Ｆｉｇ.３㊀ＣｕｒｒｅｎｔｃｏｌｌａｐｓｅｆｏｒｓａｍｐｌｅＡｗｉｔｈＮ２ｐｌａｓｍａａｎｄｓａｍｐｌｅＢｗｉｔｈＮＨ３ｐｌａｓｍａ１０ꎬ５０ꎬ１００ꎬ２００ꎬ３００Ｖ下的电流崩塌ꎮ从图３中可以看到在不同的漏极偏压下ꎬＮ２等离子体预处理器件的电流崩塌因子明显较ＮＨ３等离子体预处理的小ꎬＮ２等离子体预处理器件在偏压１００Ｖ时崩塌因子最大值为３５.６ꎬＮＨ３等离子体预处理器件为５７.５ꎻ在偏压３００Ｖ时ꎬＮＨ３等离子体预处理器件的崩塌因子最大值为８５.３ꎬＮ２等离子体预处理器件为１９.１ꎮ对比器件的动静态性能ꎬ采用Ｎ２等离子体预处理能够有效地提高器件的动态性能ꎮ３.２㊀界面预处理时间的影响图４给出了不同预处理时间下ꎬ器件转移输出特性对比ꎮ结果显示不同预处理时间对样品的基本电学性能影响不明显ꎬ预处理后器件的静态性能没有大的提高ꎮ采用ｐｕｌｓｅ￣ＤＣ表征器件的动态性能ꎮ器件测试脉冲是(５ｍｓꎬ３ｍｓ)ꎬ即关态偏压施加的时间是３ｍｓꎬ测试周期是５ｍｓꎬ器件关态偏压为(ＶＤ:５０ＶꎬＶＧＳ:－２０Ｖ)ꎮ图５中展示了不同时间预处理器件的直流/脉冲输出电流曲线对比ꎮ相比于静态输出电流ꎬＣ㊁Ｄ㊁Ｅ样品的脉冲输出电流都发生了明显下降ꎬ其中未经过Ｎ２等离子体预处理的样品Ｃ下降最为严重ꎬ预处理时间１０ｍｉｎ的样品Ｄ结果最好ꎬ样品Ｃ㊁Ｄ及样品Ｅ的饱和电流下降幅度分别为３０６.１ꎬ９９.１ꎬ１８４.５ｍＡ/ｍｍꎮ该结果表明利用Ｎ２等离子体预处理能够明显地减小器件界面导致的性能退化ꎮ对比预处理１０ｍｉｎ的样品Ｄ和处理３０ｍｉｎ的样品Ｅ的结果ꎬ发现长时间的预处理对器件的性能有一定的损害ꎬ主要原因是长时间的预处理导致表面有正电荷或者新的施主态的积累ꎬ使得器件动态性能下降[１８]ꎮ９１８㊀发㊀㊀光㊀㊀学㊀㊀报第４０卷V GS /V600500-15I D /（m A ·m m -1）400300200CD E 1000-20-10-505V d :10V20406080G m /（m S ·m m -1）（a ）V D /V6005004I D /（m A ·m m -1）400300200C D E100001081214V GS （b ）-14~2V 622V -2V-6V-10V 图４㊀不同预处理时间下器件转移输出特性曲线ꎮ(ａ)转移曲线ꎻ(ｂ)输出曲线ꎮＦｉｇ.４㊀Ｔｒａｎｓｆｅｒａｎｄｏｕｔｐｕｔｃｕｒｖｅｓｆｏｒｔｈｒｅｅｓａｍｐｌｅｓ.(ａ)Ｔｒａｎｓｆｅｒｃｕｒｖｅｓ.(ｂ)Ｏｕｔｐｕｔｃｕｒｖｅｓ.6005002V D /VI D /（m A ·m m -1）（a ）Pulse:(5ms,3ms)Based:(V d ,V gs )(50V,-20V)DC:V g :-14~2V step:4V V d :0~10VDCPulse40030020010000468101214166005002V D /VI D /（m A ·m m -1）（b ）Pulse:(5ms,3ms)Based:(V d ,V gs )(50V,-20V)DC:V gs :-14~2V step:4V V d :0~10VDC Pulse40030020010000468101214166005002V D /VI D /（m A ·m m -1）（c ）Pulse:(5ms,3ms)Based:(V d ,V gs )(50V,-20V)DC:V gs :-14~2V step:4V V d :0~10VDC Pulse 4003002001000046810121416600500CV D /VI D /（m A ·m m -1）（d ）400300D E200DCPulse100图５㊀直流㊁脉冲输出曲线对比ꎮ(ａ)样品Ｃꎻ(ｂ)样品Ｄꎻ(ｃ)样品Ｅꎻ(ｄ)实验样品直流/脉冲下饱和电流对比ꎮＦｉｇ.５㊀ＣｏｍｐａｒｉｓｉｏｎｏｆｐｕｌｓｅｄＩ￣Ｖｃｈａｒａｃｔｅｒｉｓｔｉｃｓ.(ａ)ＳａｍｐｌｅＣ.(ｂ)ＳａｍｐｌｅＤ.(ｃ)ＳａｍｐｌｅＥ.(ｄ)ＣｏｍｐａｒｉｓｏｎｏｆｓａｔｕｒａｔｉｏｎｏｕｔｐｕｔｃｕｒｒｅｎｔｄｅｎｓｉｔｙｂｅｔｗｅｅｎｐｕｌｓｅｄａｎｄＤＣ.３.３㊀界面栅介质插入层的影响图６展示了器件的转移输出特性对比ꎮ为了更明显地显示ꎬ将样品Ｆ㊁Ｇ的对比结果显示于图６(ａ)㊁(ｂ)ꎬ将样品Ｇ㊁Ｈ的对比结果显示于图６(ｃ)㊁(ｄ)ꎮ样品Ｆ㊁Ｇ和Ｈ阈值电压分别为－６.４６ꎬ－７.６２ꎬ－７.０４Ｖꎬ由此看出采用ＡｌＮ栅介质插入层导致了器件的阈值向负漂移ꎬ是因为引入ＡｌＮ插入层会在表面形成极化正电荷ꎬ影响阈值电压ꎮ图６中给出了样品Ｆ㊁Ｇ和Ｈ导通电阻分别为１３.８ꎬ１５.７ꎬ２０.６Ω ｍｍꎮ和样品Ｆ比较ꎬ样品Ｇ和Ｈ导通电阻增加的原因可能是引入ＡｌＮ介质插入层会造成导通电阻在一定范围内退化ꎬ从而使饱和电流下降[１９￣２０]ꎮ观察图６(ｃ)ꎬ发现样品Ｈ中ꎬ从－１５Ｖ扫到５Ｖ的正向及从５Ｖ回扫到－１５Ｖ的转移曲线回滞明显消除ꎬ而没有高温退火的样品Ｇ中回滞现象明显ꎮ图７给出了实验样品的正向阈值与负向阈值的对比ꎬ器件的阈值在回扫过程中会出现正向漂移ꎬＦ㊁Ｇ和Ｈ器件的阈值回滞ΔＶｔｈ(Ｖｔｈ负向－Ｖｔｈ正向)分别为４１１ꎬ５０６ꎬ１１１ｍＶꎮ和样品Ｆ相比ꎬ样品Ｈ的ΔＶｔｈ降低７２.９９％ꎬ可以看出采用退火后ＡｌＮ栅介质插入层界面处理的器件阈值回滞明显消除ꎬ说明由界面引起的器件性能退化得到控制ꎮ另外ꎬ未经过退火的ＡｌＮ介质插入层的界面处理的器件Ｇ㊀第７期韩㊀军ꎬ等:界面处理对ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓ器件动态特性的影响９１９㊀阈值回滞反而增大ꎬ这可能是ＡｌＮ材料中存在缺陷导致的ꎮ经过１０００ħ的退火过程的样品ＨꎬＡｌＮ材料存在重结晶过程ꎬ提高了ＡｌＮ材料质量ꎬ改善了界面质量ꎮ400-15V GS /VI D /（m A ·m m -1）（a ）30020010006Reference(F)AlN interlaye(G)V GS :-15~5VV D :15~-15V V D :10V-12-9-6-303２00４0６0８0Ｇm /（m S ·m m -1）400V D /VI D /（m A ·m m -1）（b ）3002001000Reference(F)AlN interlaye(G)２４６８１０R ON (F)=13.8赘·mm R ON (G)=15.7赘·mm400-15V GS /VI D /（m A ·m m -1）（c ）30020010006Anneal(H)AlN interlaye(G)V GS :15~5V V GS :15~-15V V D :10V-12-9-6-303２00４0６0８0Ｇm /（m S ·m m -1）400V D /VI D /（m A ·m m -1）（d ）3002001000AlN interlayer anneal(H)AlN interlaye(G)２４６８１０R ON (G)=15.7赘·mm R ON (G)=20.6赘·mmAlN interlayer(G)图６㊀样品转移㊁输出特性曲线对比ꎮ(ａ㊁ｂ)样品Ｆ㊁Ｇ对比ꎻ(ｃ㊁ｄ)样品Ｇ㊁Ｈ对比ꎮＦｉｇ.６㊀Ｃｏｍｐａｒｉｓｏｎｏｆｔｒａｎｓｆｅｒａｎｄｏｕｔｐｕｔｃｕｒｖｅｓｆｏｒｓａｍｐｌｅｓ.(ａꎬｂ)ＳａｍｐｌｅＦａｎｄｓａｍｐｌｅＧ.(ｃꎬｄ)ＳａｍｐｌｅＧａｎｄｓａｍｐｌｅＨ.-6.2FV t h /VGH506mV111mVV GS :-15~5V V GS :5~-15V411mV -6.0-6.4-6.6-6.8-7.0-7.2-7.4-7.6图７㊀样品Ｆ㊁Ｇ㊁Ｈ正回扫阈值回滞对比ꎮＦｉｇ.７㊀ＶｔｈｈｙｓｔｅｒｅｓｉｓｆｏｒｓａｍｐｌｅＦꎬｓａｍｐｌｅＧａｎｄｓａｍｐｌｅＨ.图８给出了样品Ｆ㊁Ｇ㊁Ｈ电流崩塌对比ꎮ对比样品Ｆ和Ｇ数据ꎬ可以看出未经过退火处理的ＡｌＮ插入层对器件的电流崩塌的改善不明显ꎬ这一结论同图７中器件阈值回滞变化相一致ꎮ对比样品Ｇ与Ｈ可以看出ꎬ器件的电流崩塌得到了很好的提高ꎬ９００Ｖ下电流崩塌因子由样品Ｇ中的４２.０４下降到样品Ｈ的４.７６ꎬ抑制效果明显ꎮ因此利用退火ＡｌＮ作为栅介质插入层进行界面处理ꎬ能够有效改善Ａｌ￣ＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件界面ꎬ提高界面质量ꎬ抑制电流崩塌ꎬ提高器件可靠性ꎮQuiesent drain bias/V80R D y n a m i c /R O N40080010060402002006001000AlN interlayer anneal(H)AlN interlayer(G)Reference(F)图８㊀样品Ｆ㊁Ｇ㊁Ｈ电流崩塌对比ꎮＦｉｇ.８㊀ＣｕｒｒｅｎｔｃｏｌｌａｐｓｅｆｏｒｓａｍｐｌｅＦꎬｓａｍｐｌｅＧａｎｄｓａｍｐｌｅＨ.４㊀结㊀㊀论本文研究了ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件制备过程中不同界面处理对其性能的影响ꎮ研究发现ꎬ经过Ｎ２等离子体预处理较ＮＨ３等离子体预处理能够降低器件的电流崩塌因子ꎬ提高器件的可靠性ꎬ在该研究基础上优化了Ｎ２等离子体预处理时间ꎬ实验结果显示１０ｍｉｎ等离子体预处理能９２０㊀发㊀㊀光㊀㊀学㊀㊀报第４０卷够有效地提高器件脉冲下电流ꎮ进一步引入ＡｌＮ栅介质插入层ꎬ实验发现利用ＡｌＮ插入层及退火工艺能够有效地改善ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴ器件界面质量ꎬ抑制电流崩塌ꎬ提高器件可靠性ꎬ器件的阈值回滞从４１１ｍＶ减小至１１１ｍＶꎬ实现在关态应力９００Ｖ下将器件的电流崩塌因子由４２.０４下降到４.７６ꎮ参㊀考㊀文㊀献:[１]ＺＨＡＮＧＺＬꎬＹＵＧＨꎬＺＨＡＮＧＸＤꎬｅｔａｌ..Ｓｔｕｄｉｅｓｏｎｈｉｇｈ￣ｖｏｌｔａｇｅＧａＮ￣ｏｎ￣ＳｉＭＩＳ￣ＨＥＭＴｓｕｓｉｎｇＬＰＣＶＤＳｉ３Ｎ４ａｓｇａｔｅｄｉｅｌｅｃｔｒｉｃａｎｄｐａｓｓｉｖａｔｉｏｎｌａｙｅｒ[Ｊ].ＩＥＥＥＴｒａｎｓ.ＥｌｅｃｔｒｏｎＤｅｖ.ꎬ２０１６ꎬ６３(２):７３１￣７３８.[２]ＬＩＵＳＣꎬＣＨＥＮＢＹꎬＬＩＮＹＣꎬｅｔａｌ..ＧａＮＭＩＳ￣ＨＥＭＴｓｗｉｔｈｎｉｔｒｏｇｅｎｐａｓｓｉｖａｔｉｏｎｆｏｒｐｏｗｅｒｄｅｖｉｃｅａｐｐｌｉｃａｔｉｏｎｓ[Ｊ].ＩＥＥＥＥｌｅｃｔｒｏｎＤｅｖ.Ｌｅｔｔ.ꎬ２０１４ꎬ３５(１０):１００１￣１００３.[３]ＫＥＬＥＫÇＩÖꎬＴＡŞＬＩＰＴꎬＹＵＨＢꎬｅｔａｌ..ＥｌｅｃｔｒｏｎｔｒａｎｓｐｏｒｔｐｒｏｐｅｒｔｉｅｓｉｎＡｌ０.２５Ｇａ０.７５Ｎ/ＡｌＮ/ＧａＮｈｅｔｅｒｏｓｔｒｕｃｔｕｒｅｓｗｉｔｈｄｉｆｆｅｒｅｎｔＩｎＧａＮｂａｃｋｂａｒｒｉｅｒｌａｙｅｒｓａｎｄＧａＮｃｈａｎｎｅｌｔｈｉｃｋｎｅｓｓｅｓｇｒｏｗｎｂｙＭＯＣＶＤ[Ｊ].Ｐｈｙｓ.ＳｔａｔｕｓＳｏｌｉｄｉꎬ２０１２ꎬ２０９(３):４３４￣４３８.[４]王凯ꎬ邢艳辉ꎬ韩军ꎬ等.掺Ｆｅ高阻ＧａＮ缓冲层特性及其对ＡｌＧａＮ/ＧａＮ高电子迁移率晶体管器件的影响研究[Ｊ].物理学报ꎬ２０１６ꎬ６５(１):０１６８０２￣１￣６.ＷＡＮＧＫꎬＸＩＮＧＹＨꎬＨＡＮＪꎬｅｔａｌ..ＧｒｏｗｔｈｓｏｆＦｅ￣ｄｏｐｅｄＧａＮｈｉｇｈ￣ｒｅｓｉｓｔｉｖｉｔｙｂｕｆｆｅｒｌａｙｅｒｓｆｏｒＡｌＧａＮ/ＧａＮｈｉｇｈｅｌｅｃｔｒｏｎｍｏｂｉｌｉｔｙｔｒａｎｓｉｓｔｏｒｄｅｖｉｃｅｓ[Ｊ].ＡｃｔａＰｈｙｓ.Ｓｉｎｉｃａꎬ２０１６ꎬ６５(１):０１６８０２￣１￣６.(ｉｎＣｈｉｎｅｓｅ)[５]ＥＬＬＥＲＢＳꎬＹＡＮＧＪＬꎬＮＥＭＡＮＩＣＨＲＪ.ＥｌｅｃｔｒｏｎｉｃｓｕｒｆａｃｅａｎｄｄｉｅｌｅｃｔｒｉｃｉｎｔｅｒｆａｃｅｓｔａｔｅｓｏｎＧａＮａｎｄＡｌＧａＮ[Ｊ].Ｊ.Ｖａｃ.Ｓｃｉ.Ｔｅｃｈｎｏｌ.Ａꎬ２０１３ꎬ３１(５):０５０８０７￣１￣２９.[６]ＸＩＯＮＧＣꎬＬＩＵＳＨꎬＬＩＹＨꎬｅｔａｌ..Ｈｏｔｃａｒｒｉｅｒｅｆｆｅｃｔｏｎｔｈｅｂｉｐｏｌａｒｔｒａｎｓｉｓｔｏｒｓ ｒｅｓｐｏｎｓｅｔｏｅｌｅｃｔｒｏｍａｇｎｅｔｉｃｉｎｔｅｒｆｅｒｅｎｃｅ[Ｊ].Ｍｉｃｒｏｅｌｅｃｔｒ.Ｒｅｌｉａｂｉｌ.ꎬ２０１５ꎬ５５(３￣４):５１４￣５１９.[７]ＺＨＡＮＧＺＬꎬＦＵＫꎬＤＥＮＧＸＧꎬｅｔａｌ..ＮｏｒｍａｌｌｙｏｆｆＡｌＧａＮ/ＧａＮＭＩＳ￣ｈｉｇｈ￣ｅｌｅｃｔｒｏｎｍｏｂｉｌｉｔｙｔｒａｎｓｉｓｔｏｒｓｆａｂｒｉｃａｔｅｄｂｙｕｓｉｎｇｌｏｗｐｒｅｓｓｕｒｅｃｈｅｍｉｃａｌｖａｐｏｒｄｅｐｏｓｉｔｉｏｎＳｉ３Ｎ４ｇａｔｅｄｉｅｌｅｃｔｒｉｃａｎｄｓｔａｎｄａｒｄｆｌｕｏｒｉｎｅｉｏｎｉｍｐｌａｎｔａｔｉｏｎ[Ｊ].ＩＥＥＥＥｌｅｃｔｒｏｎＤｅｖ.Ｌｅｔｔ.ꎬ２０１５ꎬ３６(１１):１１２８￣１１３１.[８]ＰＥＮＧＭＺꎬＺＨＥＮＧＹＫꎬＷＥＩＫꎬｅｔａｌ..Ｐｏｓｔ￣ｐｒｏｃｅｓｓｔｈｅｒｍａｌｔｒｅａｔｍｅｎｔｆｏｒｍｉｃｒｏｗａｖｅｐｏｗｅｒｉｍｐｒｏｖｅｍｅｎｔｏｆＡｌＧａＮ/ＧａＮＨＥＭＴｓ[Ｊ].Ｍｉｃｒｏｅｌｅｃｔｒ.Ｅｎｇ.ꎬ２０１０ꎬ８７(１２):２６３８￣２６４１.[９]ＶＡＮＫＯＧꎬＬＡＬＩＮＳＫＹ'ＴꎬＨＡŠＣ㊅ÍＫＳꎬｅｔａｌ..ＩｍｐａｃｔｏｆＳＦ６ｐｌａｓｍａｔｒｅａｔｍｅｎｔｏｎｐｅｒｆｏｒｍａｎｃｅｏｆＡｌＧａＮ/ＧａＮＨＥＭＴ[Ｊ].Ｖａｃｕｕｍꎬ２００９ꎬ８４(１):２３５￣２３７.[１０]ＭＥＹＥＲＤＪꎬＦＬＥＭＩＳＨＪＲꎬＲＥＤＷＩＮＧＪＭ.ＳＦ６/Ｏ２ｐｌａｓｍａｅｆｆｅｃｔｓｏｎｓｉｌｉｃｏｎｎｉｔｒｉｄｅｐａｓｓｉｖａｔｉｏｎｏｆＡｌＧａＮ/ＧａＮｈｉｇｈｅｌｅｃｔｒｏｎｍｏｂｉｌｉｔｙｔｒａｎｓｉｓｔｏｒｓ[Ｊ].Ａｐｐｌ.Ｐｈｙｓ.Ｌｅｔｔ.ꎬ２００６ꎬ８９(２２):２２３５２３￣１￣３.[１１]ＷＡＴＡＮＡＢＥＴꎬＴＥＲＡＭＯＴＯＡꎬＮＡＫＡＯＹꎬｅｔａｌ..Ｌｏｗ￣ｉｎｔｅｒｆａｃｅ￣ｔｒａｐ￣ｄｅｎｓｉｔｙａｎｄｈｉｇｈ￣ｂｒｅａｋｄｏｗｎ￣ｅｌｅｃｔｒｉｃ￣ｆｉｅｌｄＳｉＮＦｉｌｍｓｏｎＧａＮｆｏｒｍｅｄｂｙｐｌａｓｍａｐｒｅｔｒｅａｔｍｅｎｔｕｓｉｎｇｍｉｃｒｏｗａｖｅ￣ｅｘｃｉｔｅｄｐｌａｓｍａ￣ｅｎｈａｎｃｅｄｃｈｅｍｉｃａｌｖａｐｏｒｄｅｐｏｓｉｔｉｏｎ[Ｊ].ＩＥＥＥＴｒａｎｓ.ＥｌｅｃｔｒｏｎＤｅｖ.ꎬ２０１３ꎬ６０(６):１９１６￣１９２２.[１２]ＫＩＭＪＨꎬＣＨＯＩＨＧꎬＨＡＭＷꎬｅｔａｌ..Ｅｆｆｅｃｔｓｏｆｎｉｔｒｉｄｅ￣ｂａｓｅｄｐｌａｓｍａｐｒｅｔｒｅａｔｍｅｎｔｐｒｉｏｒｔｏＳｉＮｘＰａｓｓｉｖａｔｉｏｎｉｎＡｌＧａＮ/ＧａＮｈｉｇｈ￣ｅｌｅｃｔｒｏｎ￣ｍｏｂｉｌｉｔｙｔｒａｎｓｉｓｔｏｒｓｏｎｓｉｌｉｃｏｎｓｕｂｓｔｒａｔｅｓ[Ｊ].Ｊｐｎ.Ｊ.Ａｐｐｌ.Ｐｈｙｓ.ꎬ２０１０ꎬ４９(４Ｓ):０４ＤＦ０５. [１３]ＨＵＡＮＧＳꎬＪＩＡＮＧＱＭꎬＹＡＮＧＳꎬｅｔａｌ..ＥｆｆｅｃｔｉｖｅｐａｓｓｉｖａｔｉｏｎｏｆＡｌＧａＮ/ＧａＮＨＥＭＴｓｂｙＡＬＤ￣ｇｒｏｗｎＡｌＮｔｈｉｎｆｉｌｍ[Ｊ].ＩＥＥＥＥｌｅｃｔｒｏｎＤｅｖ.Ｌｅｔｔ.ꎬ２０１２ꎬ３３(４):５１６￣５１８.[１４]ＥＤＷＡＲＤＳＡＰꎬＭＩＴＴＥＲＥＤＥＲＪＡꎬＢＩＮＡＲＩＳＣꎬｅｔａｌ..ＩｍｐｒｏｖｅｄｒｅｌｉａｂｉｌｉｔｙｏｆＡｌＧａＮ￣ＧａＮＨＥＭＴｓｕｓｉｎｇａｎＮＨ３/ｐｌａｓ￣ｍａｔｒｅａｔｍｅｎｔｐｒｉｏｒｔｏＳｉＮｐａｓｓｉｖａｔｉｏｎ[Ｊ].ＩＥＥＥＥｌｅｃｔｒｏｎＤｅｖ.Ｌｅｔｔ.ꎬ２００５ꎬ２６(４):２２５￣２２７.[１５]ＨＡＳＨＩＺＵＭＥＴꎬＯＯＴＯＭＯＳꎬＯＹＡＭＡＳꎬｅｔａｌ..ＣｈｅｍｉｓｔｒｙａｎｄｅｌｅｃｔｒｉｃａｌｐｒｏｐｅｒｔｉｅｓｏｆｓｕｒｆａｃｅｓｏｆＧａＮａｎｄＧａＮ/ＡｌＧａＮｈｅｔｅｒｏｓｔｒｕｃｔｕｒｅｓ[Ｊ].Ｊ.Ｖａｃ.Ｓｃｉ.Ｔｅｃｈｎｏｌ.ꎬ２００１ꎬ１９(４):１６７５￣１６８１.[１６]ＲＯＭＥＲＯＭＦꎬＪＩＭÉＮＥＺＪＩＭＥＮＥＺＡꎬＭＩＧＵＥＬ￣ＳÁＮＣＨＥＺＭＩＧＵＥＬ￣ＳＡＮＣＨＥＺＪꎬｅｔａｌ..ＥｆｆｅｃｔｓｏｆＮ２ｐｌａｓｍａｐｒｅｔｒｅａｔ￣ｍｅｎｔｏｎｔｈｅＳｉＮｐａｓｓｉｖａｔｉｏｎｏｆＡｌＧａＮ/ＧａＮＨＥＭＴ[Ｊ].ＩＥＥＥＥｌｅｃｔｒｏｎＤｅｖ.Ｌｅｔｔ.ꎬ２００８ꎬ２９(３):２０９￣２１１.[１７]ＨＡＳＨＩＺＵＭＥＴꎬＯＯＴＯＭＯＳꎬＩＮＡＧＡＫＩＴꎬｅｔａｌ..ＳｕｒｆａｃｅｐａｓｓｉｖａｔｉｏｎｏｆＧａＮａｎｄＧａＮ/ＡｌＧａＮｈｅｔｅｒｏｓｔｒｕｃｔｕｒｅｓｂｙｄｉｅｌｅｃ￣ｔｒｉｃｆｉｌｍｓａｎｄｉｔｓａｐｐｌｉｃａｔｉｏｎｔｏｉｎｓｕｌａｔｅｄ￣ｇａｔｅｈｅｔｅｒｏｓｔｒｕｃｔｕｒｅｔｒａｎｓｉｓｔｏｒｓ[Ｊ].Ｊ.Ｖａｃ.Ｓｃｉ.Ｔｅｃｈｎｏｌ.Ｂꎬ２００３ꎬ２１(４):１８２８￣１８３８.㊀第７期韩㊀军ꎬ等:界面处理对ＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓ器件动态特性的影响９２１㊀[１８]ＲＥＩＮＥＲＭꎬＬＡＧＧＥＲＰꎬＰＲＥＣＨＴＬＧꎬｅｔａｌ..Ｍｏｄｉｆｉｃａｔｉｏｎｏｆ ｎａｔｉｖｅ ｓｕｒｆａｃｅｄｏｎｏｒｓｔａｔｅｓｉｎＡｌＧａＮ/ＧａＮＭＩＳ￣ＨＥＭＴｓｂｙｆｌｕｏｒｉｎａｔｉｏｎ:ｐｅｒｓｐｅｃｔｉｖｅｆｏｒｄｅｆｅｃｔｅｎｇｉｎｅｅｒｉｎｇ[Ｃ].ＰｒｏｃｅｅｄｉｎｇｓｏｆＩＥＥＥＩｎｔｅｒｎａｔｉｏｎａｌＥｌｅｃｔｒｏｎＤｅｖｉｃｅｓＭｅｅｔｉｎｇꎬＷａｓｈ￣ｉｎｇｔｏｎꎬＤＣꎬＵＳＡꎬ２０１５:３５.５.１￣３５.５.４.[１９]ＡＣＵＲＩＯＥꎬＣＲＵＰＩＦꎬＭＡＧＮＯＮＥＰꎬｅｔａｌ..ＩｍｐａｃｔｏｆＡｌＮｌａｙｅｒｓａｎｄｗｉｃｈｅｄｂｅｔｗｅｅｎｔｈｅＧａＮａｎｄｔｈｅＡｌ２Ｏ３ｌａｙｅｒｓｏｎｔｈｅｐｅｒｆｏｒｍａｎｃｅａｎｄｒｅｌｉａｂｉｌｉｔｙｏｆｒｅｃｅｓｓｅｄＡｌＧａＮ/ＧａＮＭＯＳ￣ＨＥＭＴｓ[Ｊ].Ｍｉｃｒｏｅｌｅｃｔｒ.Ｅｎｇ.ꎬ２０１７ꎬ１７８:４２￣４７.[２０]ＨＵＡＮＧＳꎬＪＩＡＮＧＱＭꎬＹＡＮＧＳꎬｅｔａｌ..ＭｅｃｈａｎｉｓｍｏｆＰＥＡＬＤ￣ｇｒｏｗｎＡｌＮｐａｓｓｉｖａｔｉｏｎｆｏｒＡｌＧａＮ/ＧａＮＨＥＭＴｓ:ｃｏｍｐｅｎ￣ｓａｔｉｏｎｏｆｉｎｔｅｒｆａｃｅｔｒａｐｓｂｙｐｏｌａｒｉｚａｔｉｏｎｃｈａｒｇｅｓ[Ｊ].ＩＥＥＥＥｌｅｃｔｒｏｎＤｅｖ.Ｌｅｔｔ.ꎬ２０１３ꎬ３４(２):１９３￣１９５.韩军(１９６４－)ꎬ男ꎬ北京人ꎬ博士ꎬ副教授ꎬ２００８年于北京工业大学获得博士学位ꎬ主要从事半导体材料与器件方面的研究ꎮＥ￣ｍａｉｌ:ｈａｎｊｕｎ＠ｂｊｕｔ.ｅｄｕ.ｃｎ邢艳辉(１９７４－)ꎬ女ꎬ吉林德惠人ꎬ博士ꎬ副教授ꎬ２００８年于北京工业大学获得博士学位ꎬ主要从事氮化镓半导体材料的生长㊁测试分析及器件等方面的研究ꎮＥ￣ｍａｉｌ:ｘｉｎｇｙａｎｈｕｉ＠ｂｊｕｔ.ｅｄｕ.ｃｎ。
