面向纳米尺度金属互连线的蒙特卡洛模拟方法研究

胡远钊1 杜刚2 杨燚1 赵钰迪1,2 赵凯1,2,†

1.北京信息科技大学光电测试技术及仪器教育部重点实验室, 北京 100101; 2.北京大学信息科学技术学院, 北京 100871; †通信作者, E-mail: zhaokai@pku.edu.cn

摘要 考虑各种尺度下金属互连线中主要的电子散射机制, 提出一种可以准确地仿真金属互连线中电子输运特性的蒙特卡洛模拟方法。模拟结果表明, 等离子激元散射对体材料互连线的电阻率贡献最大, 其次是电子间散射; 晶粒间界散射则主导纳米尺度线宽的金属电阻率, 是导致“尺寸效应”的主要原因。通过与实验数据对比, 证明所提方法可以准确地模拟从体材料到纳米尺度的金属互连电阻率。

关键词 蒙特卡洛方法; 散射; 电阻率; 金属互连

自 20 世纪 80 年代后期开始, 由于铜(Cu)互连具有更低的电阻率和更好的抗电迁移特性, 因此取代铝互连, 并广泛应用于超大规模集成电路中[1–2]。近年来, 芯片后端的线宽进入纳米尺度, 铜互连线的电阻率随着线宽的缩小而迅速增加[2–4], 即“尺寸效应”[5–6], 从而增加了电路的互连 RC 延迟[7]。同时, 小尺寸线宽的铜互连线的电迁移现象显著地降低了器件的可靠性[8–9]。这些问题制约了器件性能的发展, 因此寻找铜互连的替代材料成为一个紧迫的重要任务。与铜相比, 钴(Co)和钌(Ru)具有更小的平均电子自由程及更高的熔点, 具备更好的抗“尺寸效应”和抗电迁移能力[9–13]。这些优点使 Co和 Ru 成为下一代互连材料的选择方案。

蒙特卡洛(Monte Carlo)方法是一种利用统计的方法求解数学物理方程的数值方法。用此方法间接求解电子的玻尔兹曼方程, 可以直观地模拟各种散射过程, 研究不同散射机制对材料电阻率的贡献。蒙特卡洛方法在半导体领域的应用可以追溯到 20世纪 60 年代, Kurosawa[14]首次利用此方法研究外加电场下半导体材料中自由电子的输运问题, 并利用统计方法得到稳态的电子分布函数。20 世纪 70 年代, 非抛物线能带修正的蒙特卡洛方法开始得到完善, 广泛应用于 GaAs, Si 以及 Ge 等材料的载流子输运特性的研究中[15–16]。随着器件特征尺寸的缩小以及输运方向电场的增大, 需要考虑的载流子能量范围也不断地增加, 全能带的蒙特卡洛方法逐渐发展起来, 以便适应这种需求[17–22]

在小尺寸金属互连中, 导致“尺寸效应”的主要因素是晶粒大小和线边缘粗糙度, 即电子在输运过程中经历的晶粒间界散射和表面粗糙散射[6]。为了研究小尺寸情况下金属互连线电阻率升高的物理机制, 定量地分析晶粒间界散射和表面粗糙散射的影响, 本文提出一种基于蒙特卡洛方法的金属互连线输运特性模拟方法, 模拟电子经历的自由飞行和散射这两个物理过程。利用该方法模拟体材料和几十纳米尺度下金属互连线中电子的输运过程, 以此检验该方法的收敛性和准确性。

1 模拟方法与验证

为了探寻不同线宽情况下电阻率变化的内在物理机制以及不同制备工艺和条件对材料电阻率的影响, 开展金属互连线中的电子输运特性的研究具有很强的理论指导意义。在几十纳米及更宽的尺度下, 包含散射的非平衡格林函数方法过于耗费计算时间[23], 准经典理论框架下的玻尔兹曼方程仍然是描述金属材料电子输运特性最合适的途径[24]。电子的玻尔兹曼输运方程求解主要有直接数值求解和蒙特卡洛方法两种方式, 其中直接数值求解方法面对的是复杂的高维微分积分方程, 在处理散射项的卷积时往往面临一些挑战; 蒙特卡洛方法将电子的输运过程分为交替进行的自由飞行和随机散射两部分[24], 通过对电子的自由飞行和散射的大量模拟和统计, 实现间接求解电子的玻尔兹曼方程的目的。

1.1 金属互连线的单粒子蒙特卡洛模拟方法

理论上, 蒙特卡洛模拟的粒子个数和模拟的步数趋于无限大时, 就可以得到玻尔兹曼输运方程精确解。为了在计算时间与精度之间取得平衡, 程序会设置一个总时间步长作为上限。在金属互连线中, 电场分布是恒定的, 所以求解区域内不需要求解泊松方程, 可以采用单粒子蒙特卡洛方法[25]进行简化的计算, 也就是在每个时间步长 Δt 内, 让 N 个电子依次进行自由飞行和散射, 并将每个电子在该Δt 结束时的能量终态 ϵ'和动量 k'作为下一个 Δt 内每个电子的初始状态。如此循环往复, 直至模拟结果收敛或总步长达到上限, 模拟过程结束。

单粒子蒙特卡洛方法中, 每个时间步长的核心在于: 1)计算和更新自由飞行时间; 2)选择散射机制; 3)计算电子散射终态。这三者构成一个电子输运过程中基本的循环单元。程序收敛后, 统计所有电子的波矢和能量, 就可以得到电子的分布函数和宏观物理量。

金属互连线中影响电阻率的电子散射机制有 6种, 分别为声学波形变势散射(acoustic phonon scattering, APS)、光学波形变势散射(optical phonon scattering, OPS)、电子对电子散射(electron to elec-tron scattering, ETES), 等离子激元散射(plasma ex-cimer scattering, PES)、晶粒间界散射(grain boun-dary scattering, GBS)以及表面粗糙散射(surface roughness scattering, SRS)。为了简化计算, 还引入自散射(self-scattering, SS)这一虚拟的散射机制。为了计算各种不同散射机制的散射率, 最重要的是确定每种散射机制的散射势函数。

波矢为 k 的声学波形变势散射的散射势[25]

width=186,height=28.5 (1)

式中, aq 代表振幅, width=12,height=11.25是形变势因子, 具有能量的量纲, e 是沿振动方向的单位矢量, ωq 是声学波的频率, width=9.75,height=12是声学波的相位。

波矢为 k 的光学波形变势散射的散射势公式[25]如下:

width=186,height=28.5 (2)

式中, D0 为光学波形变势常数。

金属材料中, 电子间散射类似于半导体中的电离杂质散射, 电子间的屏蔽库仑势[26–27]可表示为

width=111,height=30.75 (3)

式中, n 为电子浓度, εh 为高频介电常数, e 为单位电荷量, LD 为德拜长度。

在金属中, 电子之间的库伦相互作用使电子密度分布呈现一定的波动性, 该过程类似晶格振动波产生的振荡表现出来的过程, 所以振荡能量的变化是量子化的, 只能是 ωq 的整数倍。因此, 用 n 个能量为 ωq, 动量为 q 的准粒子代表量子数为 n, 波矢为 q 的振荡, 当准粒子发生振荡时, 等离子激元散射的散射势计算公式[28]

width=159.7,height=32.25 (4)

在大尺度的互连线中, 晶粒尺寸 a 与晶界厚度b 的比率 a/b 超过两个数量级[29], 电子能量在渡越晶粒间界时的损耗可忽略不计, 对输运过程的影响很小。在小尺寸互连线中, 晶粒的尺寸随着线宽的缩小而减小(图 1(a)), 并在纳米尺度的线宽下演化成“竹节”状分布(图 1(b))。电子在这种竹节结构的晶粒间穿行需要频繁克服晶粒间界的势垒[30]。本文用 sinc 函数对势垒的势函数进行近似(图 1 (c)), 并据此使用多个势函数的叠加形式。定义晶粒间界散射的散射势为

width=116.25,height=20.25 (5)

其中, P 为势垒的高度, xi=ia+(i–1)⋅b 为第 i 个晶粒间界的位置。

互连线中运动的电子会与粗糙表面发生频繁散射, 导致动量弛豫[31]。镜面反射系数 μ 和粗糙系数σ 分别表示表面的反射情况和表面的粗糙程度, 其中 μ 的取值范围为[0, 1], μ = 0 表示完全漫反射, μ = 1表示完全镜面反射; σ 的取值范围为(0, 5], σ 值越大, 表面越平滑。在本文的小尺寸互连线模拟中, μσ分别设置为 0 和 1.5, 表面粗糙散射的散射势取为

width=135,height=29.25 (6)

在模拟过程中, 应按照不同的物理结构选择适当的散射机制, 但其中必须包含至少一种非弹性散射, 确保能量平衡[25]。在本文方法中, 光学波形变势散射、等离子激元散射和晶粒间界散射为非弹性散射。在势函数的基础上, 可以根据费米黄金定则[32], 进一步计算得到各种散射机制对应的散射率 τ(t)。

在程序中, 每次计算自由飞行时间会消耗大量计算资源和时间。引入“自散射”概念[33]可以解决这个问题。自散射发生时, 电子的波矢、能量和分布函数均不发生改变。如果令包括自散射在内的散射率之和为常数, 则自由飞行时间的计算可以由求解一个积分问题简化为一个包含随机数的代数方程, 大幅缩短计算时间。

width=229.6,height=135.45

图1 40 nm互连线中晶粒间界与势函数示意图[29]

Fig. 1 Schematic diagram of grain boundary and its potential function in a 40 nm line[29]

表 1 给出模拟中 Cu、Ru 和 Co 的散射机制的各项物理参数。

1.2 模拟方法的收敛性与结果验证

通过分别模拟 10, 40 和 200nm 三个不同尺寸的Cu, Ru 和 Co 互连线的电阻率来验证方法的准确性, 并给出蒙特卡罗方法在模拟金属材料时的收敛特性。考虑到蒙特卡洛方法难以消除数值涨落, 所以本文取 n 个时间步长 Δt 内的平均电阻率width=15.05,height=17.55作为收敛的判据, 当width=41.95,height=17.55<105 时, 认为程序已经收敛。这样, 就可以避免由程序中随机数引起的随机涨落的干扰。

金属体材料中晶粒尺寸较大, 晶界厚度约为 1 nm 数量级[29]。在外场作用下, 内部电子具有较大的漂移速度, 电子在穿越不同晶粒间界的能量损失可忽略不计。因此, 晶粒间界的势垒作用几乎不会对电子的输运产生影响, 此时线宽尺度远高于材料的平均电子自由程, 所以表面粗糙散射的影响也可以忽略不计。可以认为, 200nm 线宽已经满足体材料条件[37], 所以在 200nm 模拟中只考虑除晶粒间界散射和表面粗糙散射外的散射机制, 而 10nm 和 40nm线宽考虑了所有的散射机制。

本文中的初始参数设置为 Ttotal=2×104, N=5× 105,TF =1×10–13s, 电场强度 E 为 1kV/cm。图 2 给出仿真过程中 Ru, Co 和 Cu 三种材料的电阻率随时间步长的变化, 其中灰色分布散点为瞬时电阻率width=15.05,height=15.05width=15.05,height=17.55为最近 500 个步长的平均电阻率。从图 2 可以看出, 模拟初期, 由于电子经历的散射次数较少, 电阻率值波动较为明显; 随着时间步长增加, 散射次数的增加使得程序逐渐收敛。图 3 对比 3 种材料在不同线宽下的收敛特性, 可以看出, 大线宽收敛速度较慢, 而小尺寸下收敛非常快。这是因为在给定外场下, 电子发生散射的概率随线宽的减小而增加, 使得电子遍历 k 空间各种状态的速度更快, 从而加速收敛。

表1 Cu, Ru和Co的散射参数

Table 1 Scattering parameters of Cu, Ru and Co

参数CuRuCo 声学波形变势因子Ξ (eV)6.107.594.31 光学波形变势常数D0 (109 eV/cm)1.000.700.65 电子浓度n (1028/m3)[34–36]8.487.2118.938 晶粒间界势垒高度P(eV)0.100.10.1 镜面反射系数000 粗糙系数1.51.51.5

width=220.8,height=232.8

图2 Cu, Ru和Co的体材料电阻率随模拟时间的变化

Fig. 2 Convergence behavior of bulk resistivity with the increase of time steps

本文的模拟是利用普通 PC 机完成, 每个模拟任务对内存的需求约为 4GB。在使用型号为 AMD RyzenTM 4800H 的 CPU 进行计算时, 每 400 个时间步长需要约一小时计算时间, 其中散射机制的选择消耗了最多的计算时间。

表 2 给出 1kV/cm 和 5kV/cm 两种外场条件下, 3 种不同材料的收敛速度, 可以看出, 低场下电子平均漂移速度小, 动量弛豫时间较长, 在统计结果中表现为width=15.05,height=16.9的涨落更明显, 收敛时间更长。表 3对比 3 种金属的体材料电阻率与实验数据[38], 可以看出, 3 种材料的模拟结果都与实验数据有非常好的吻合, 验证了本方法的准确性。

2 方法验证与仿真结果

本文以 200nm 线宽为例, 研究体材料中 4 种主要散射机制随能量的变化以及材料电阻率随温度的变化, 并研究电阻率随着线宽减小的变化规律, 指出晶粒间界和表面粗糙散射在纳米尺度下的显著增强是产生“尺寸效应”的根源。

width=212.25,height=485.5

图3 在1 kV/cm电场下不同尺寸的收敛情况

Fig. 3 Convergence behavior at 1 kV/cm E-field with different linewidths

表2 不同电场下蒙特卡洛程序的收敛时间(10–10s)

Table 2 Convergence behavior of MC program with different electric fields (10–10s)

材料1 kV/cm 5 kV/cm 10 nm40 nm200 nm10 nm40 nm200 nm Cu46.5934.57 Ru57.5114.56.58.5 Co4.5710.545.58

表3 体材料电阻率的模拟值与实验值的对比(μΩ·cm)

Table 3 Bulk resistivity comparison between simulation and experiment results (μΩ·cm)

材料仿真数据实验数据 Cu1.6861.68 Co6.1746.18 Ru7.7617.8

2.1 金属体材料的散射率与电阻率

在体材料中, 由于晶粒的尺寸较大, 所以不需要考虑晶粒间界散射。并且, 电子的平均自由程远小于线宽, 所以发生表面粗糙散射的概率也比较小, 从而不需要考虑表面粗糙散射。因此, 体材料中的主要散射机制为声学、光学声子散射、电子间散射和等离子激元散射。为研究体材料中 4 种散射机制对电阻率的贡献, 本文将所有散射率之和进行归一化, 结果如图 4 所示。这样, 就可以利用随机数 r 来选择散射机制: 当 r 落在图中最近邻曲线之下的区域时, 说明选择了该曲线代表的散射机制; 当 r 落在图中最上方(代表等离子激元散射的曲线的上方)区域时, 则表明发生了一次自散射。可以看出, 等离子激元散射率占比最大, 是几种金属体材料中发生频率最高的散射机制, 对电子输运行为的影响最显著。电子间散射对 Cu 和 Ru 材料电阻率的贡献也较大, 对 Co 电阻率的贡献非常小。然而, 与等离子激元散射相比, 都小了一个数量级以上。

width=198.1,height=458.85

图4 金属体材料的归一化散射率随能量的变化规律

Fig. 4 Normalized scattering rate with respect of energy for bulk metal interconnects

width=219.95,height=184.4

图5 体材料电阻率的模拟结果、经验公式与实验结果的对比

Fig. 5 Resistivity comparison of simulated results, empirical formula and measured data

width=216.1,height=179

图6 纳米尺度互连线电阻率随线宽的变化

Fig. 6 Resistivity of nanoscale interconnects with different width

通常, 体材料金属的电阻率可以用如下经验公式计算:

width=98.3,height=15.05 (7)

其中, ρ273K为 273K 时的材料电阻率; α 为温度系数, 其中 Cu 为 4.336×10–3[39]、Co 为 6.944×10–3[40], Ru为 4.065×10–3[41]。图 5 对比该经验公式与本文蒙特卡洛模拟数据和实验测试结果[38,42]随温度的变化情况, 可以看出, Cu 的模拟结果与实验最吻合, 经验公式也比较接近实验和模拟结果; Ru 和 Co 的模拟结果与实验结果的一致性也非常好, 经验公式则有较大的误差。上述对比结果说明, 本文提出的蒙特卡洛方法可以准确地模拟各种温度下不同金属材料的电阻率。由于金属体材料中等离子激元散射占据主导地位, 当环境温度超过金属的德拜温度时, 等离子激元散射率正比于等离子激元的数目 nq, 而 nq正比于温度, 所以随温度升高, 体电阻率整体上呈上升的变化趋势。

width=398.15,height=450.6

图7 纳米尺度下金属中电子散射率与能量的关系

Fig. 7 Relationship between electron energy and scattering rate in nanoscale interconnects

2.2 纳米尺度互连线的电阻率变化规律

图 6 给出金属线宽从 48nm 减小至 10nm 时, 本文蒙特卡洛模拟数据与实验测试的电阻率[43–47]的对比, 可以看出, 由于制备条件不同, 相同材料的电阻率在同一线宽下存在一定程度的差异, IMEC 的10nm 线宽尤为明显[45], 与 12nm 线宽相比, 电阻率出现明显的跃升。此外, IMEC 的其他节点[45]和相关研究[44,46‒47]中的实验数据与本文模拟数据的吻合度都比较好。

模拟结果显示, Cu 的电阻率随线宽减小的增幅几乎翻倍, 是 3 种材料中最大的, 但是电阻率仍然是三者中最小的。在两种新型互连线材料中, Co 的电阻率在 30nm 以上的互连线中低于 Ru, 而进入更小的线宽后电阻率显著增加。这是由于 Co 的晶粒间界散射和表面粗糙散射两种机制随着线宽的缩小都显著增强, 尤其是在 10nm 节点中, Co 的晶粒间界和表面粗糙散射率都反超 Ru 的这两种散射率(图7(c)和(e))。

从图 7 可以看出, 3 种材料的晶粒间界散射率都随电子能量升高而降低, 说明晶粒间界的势垒作用随电子能量的增加而显著减弱。考虑到电子主要分布在低能级上, 所以晶粒间界散射的作用尤为突出。值得注意的是, Cu 的晶粒间界散射率随线宽的变化更为明显, 如图 7(a)~(c)所示。表面粗糙散射率随电子能量的升高而增加, 这是由于电子能量越高, 平均漂移速度越大, 与表面发生碰撞的概率越高。当线宽小于 Cu 材料对应的平均电子自由程时, 平均电子自由程就变成输运方向上实际的线宽长度, 由表面引起的散射会更加剧烈, 线宽收缩时散射率的增长也会更明显。Ru 和 Co 平均电子自由程较小, 所以表面粗糙散射率随着线宽缩小, 不会表现出非常明显的增长。对比图 7 中 3 种材料的散射率可以发现, 在小尺寸互连线中, 低能量区域晶粒间界散射率都远高于表面粗糙散射率和等离子激元散射率。因此, 在小尺寸互连线中, 晶粒间界散射是造成“尺寸效应”的主要原因。

3 结论

本文提出一种基于蒙特卡洛方法的金属互连线模拟方法, 通过模拟 Cu, Ru 和 Co 的电阻率验证了该方法的收敛性, 并研究 200nm 线宽的 Cu, Ru 和Co 互连线中电阻率随温度的变化情况。在此基础上, 针对未来先进工艺节点中的纳米尺度互连线中影响电阻率的因素进行了研究。结果表明, 在纳米尺度下晶粒间界散射对电阻率影响最大, 等离子激元散射和表面粗糙散射次之。在未来先进工艺节点中, 呈“竹节状”分布的纳米尺度晶粒导致频繁发生的晶粒间界散射是产生“尺寸效应”的首要原因。

参考文献

[1] Andricacos P C, Uzoh C, Dukovic J O, et al. Damas-cene copper electroplating for chip interconnections. IBM Journal of Research and Development, 1998, 42 (5): 567–574

[2] Andricacos P C. Copper On-chip Interconnections. The Electrochemical Society Interface, 1999, 8(1): 32–37

[3] Shimada M, Moriyama M, Ito K, et al. Electrical resistivity of polycrystalline Cu interconnects with nano-scale linewidth. Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures Processing, Measurement, and Phenomena, 2006, 24(1): 190–194

[4] Steinhögl W, Schindler G, Steinlesberger G, et al. Size-dependent resistivity of metallic wires in the mesoscopic range. Physical Review B, 2002, 66(7): 075414

[5] Hu C K, Rosenberg R, Lee K Y. Electromigration path in Cu thin-film lines. Applied Physics Letters, 1999, 74(20): 2945–2947

[6] Wu W, Maex K. Studies on size effect of copper interconnect lines // 6th International Conference on Solid-State and Integrated Circuit Technology (ICSICT). Piscataway, 2001: 416–418

[7] Chen F, Gardner D. Influence of line dimensions on the resistance of Cu interconnections. IEEE Electron Device Letters, 1998, 19(12): 508–510

[8] Zahedmanesh H, Pedreira O V, Wilson C, et al. Copper electromigration: prediction of scaling limits. IEEE IITC/MAM. Brussels, 2019: 1–3

[9] Gall D. The resistivity bottleneck: the search for new interconnect metals // 2020 International Symposium on VLSI Technology, Systems and Applications (VLSI- TSA). Piscataway, 2020: 112–113

[10] Auth C, Aliyarukunju A, Asoro M, et al. A 10 nm high performance and low-power CMOS technology featuring 3 rd generation FinFET transistors, self-aligned quad patterning, contact over active gate and cobalt local interconnects // 2017 IEEE International Electron Devices Meeting (IEDM). New York, 2017: 29.1.1–29.1.4

[11] Jablonka L, Riekehr L, Zhang Z, et al. Highly conductive ultrathin Co films by high-power impulse magnetron sputtering. Applied Physics Letters, 2018, 112(4): 043103

[12] Zhang X, Huang H, Patlolla R, et al. Methods to lower the resistivity of ruthenium interconnects at 7 nm node and beyond // 2017 IEEE International Interconnect Technology Conference (IITC). New York, 2017: 1–3

[13] Zhang X, Huang H, Patlolla R, et al. Ruthenium interconnect resistivity and reliability at 48 nm pitch // 2016 IEEE International Interconnect Technology Conference/Advanced Metallization Conference (IITC/ AMC). New York, 2016: 31–33

[14] Kurosawa T. Monte Carlo calculation of hot electron problems // Proceedings of the International Confe-rence on the Physics of Semiconductors. Tokyo: Phy-sical Society of Japan, 1966: 424–426

[15] Fawcett W, Boardman A D, Swain S. Monte Carlo determination of electron transport properties in gallium arsenide. Journal of Physics and Chemistry of solids, 1970, 31(9): 1963–1990

[16] Jacoboni C, Minder R, Majni G. Effects of band non-parabolicity on electron drift velocity in silicon above room temperature. Journal of Physics and Chemistry of Solids, 1975, 36(10): 1129–1133

[17] Fischetti M V, Laux S E. Monte Carlo analysis of electron transport in small semiconductor devices including band-structure and space-charge effects. Physical Review B, 1988, 38(14): 9721–9745

[18] Kunikiyo T, Takenaka M, Kamakura Y, et al. A Monte Carlo simulation of anisotropic electron transport in silicon including full band structure and anisotropic impact-ionization model. Journal of Applied Phy- sics, 1994, 75(1): 297–312

[19] Fischetti M V, Laux S E. Monte Carlo simulation of electron transport in Si: the first 20 years // Procee-dings of the 26th European Solid State Device Re-search Conference. Piscataway, 1996: 813–820

[20] Du Gang, Liu Xiaoyan, Xia Zhiliang, et al. Monte Carlo simulation of p-and n-channel GOI MOSFETs by solving the quantum Boltzmann equation. IEEE Transactions on Electron Devices, 2005, 52(10): 2258–2264

[21] Du Gang, Liu Xiaoyan, Han Ruqi. Quantum Boltz-mann equation solved by Monte Carlo method for nano-scale semiconductor devices simulation. Chinese Physics, 2006, 15(1): 177–181

[22] Lei Shen, Di Shaoyan, Yin Longxiang, et al. Impact of crystal orientation and surface scattering on DG-MOSFETs in quasi-ballistic region. Journal of Semi-conductors, 2018, 39(7): 074002

[23] Datta S. Quantum transport: atom to transistor. Cam-bridge: Cambridge University Press, 2005

[24] Jacoboni C, Reggiani L. The Monte Carlo method for the solution of charge transport in semiconductors with applications to covalent materials. Reviews of Modern Physics, 1983, 55(3): 645–705

[25] 叶良修. 小尺寸半导体器件的蒙特卡罗模拟. 北京: 科学出版社, 1997

[26] Lugli P, Ferry D K. Electron-electron interaction and high field transport in Si. Applied physics letters, 1985, 46(6): 594–596

[27] Lugli P, Ferry D K. Effect of electron-electron scat-tering on Monte Carlo studies of transport in sub-micron semiconductors devices. Physica B+C, 1983, 117: 251–253

[28] Katoh R, Kurata M, Yoshida J. Self-consistent particle simulation for (AlGa) As/GaAs HBTs with improved base-collector structures. IEEE Transactions on Elec-tron Devices, 1989, 36(5): 846–853

[29] Simoes S, Calinas R, Vieira M T, et al. In situ TEM study of grain growth in nanocrystalline copper thin films. Nanotechnology, 2010, 21(14): 145701

[30] Mayadas A F, Shatzkes M. Electrical-resistivity model for polycrystalline films: the case of arbitrary ref-lection at external surfaces. Physical review B, 1970, 1(4): 1382–1389

[31] Fuchs K. The conductivity of thin metallic films according to the electron theory of metals. Mathe-matical Proceedings of the Cambridge Philosophical Society, 1938, 34(1): 100–108

[32] Fermi E. Nuclear physics: a course given by Enrico Fermi at the University of Chicago. Chicago: Uni-versity of Chicago Press, 1950

[33] Rees H D. Calculation of distribution functions by exploiting the stability of the steady state. Journal of physics and Chemistry of solids, 1969, 30(3): 643–655

[34] Saslow W M. Electricity, magnetism, and light. Cam-bridge: Academic Press, 2002

[35] Rohlf J W. Modern Physics from alpha to Z0. New York: John Wiley, 1994

[36] Dabral A, Pourtois G, Sankaran K, et al. Study of the intrinsic limitations of the contact resistance of metal/semiconductor interfaces through atomistic simulations. ECS Journal of Solid State Science and Technology, 2018, 7(6): N73–N80

[37] Schindler G, Penka S, Steinlesberger G, et al. Re-liability studies of narrow Cu lines. Microelectronic Engineering, 2005, 82(3/4): 645–649

[38] Gall D. Electron mean free path in elemental metals. Journal of Applied Physics, 2016, 119(8): 085101

[39] Schuster C E, Vangel M G, Schafft H A. Improved estimation of the resistivity of pure copper and elec-trical determination of thin copper film dimensions. Microelectronics Reliability, 2001, 41(2): 239–252

[40] White G K, Woods S B. Low temperature resistivity of the transition elements: cobalt, tungsten, and rhenium. Canadian Journal of Physics, 1957, 35(5): 656–665

[41] White G K, Woods S B. Low temperature resistivity of the transition elements: ruthenium and osmium. Canadian Journal of Physics, 1958, 36(7): 875–883

[42] De Vries J W C. Temperature and thickness depen-dence of the resistivity of thin polycrystalline alu-minium, cobalt, nickel, palladium, silver and gold films. Thin Solid Films, 1988, 167(1/2): 25–32

[43] VanOlmen J, List S, Tokei Z, et al. Cu resistivity scaling limits for 20nm copper damascene lines // 2007 IEEE International Interconnect Technology Conference. Piscataway, 2007: 49–51

[44] Chawla J S, Gstrein F, O’Brien K P, et al. Electron scattering at surfaces and grain boundaries in Cu thin films and wires. Physical Review B, 2011, 84(23): 235423

[45] Van der Veen M H, Heyler N, Pedreira O V, et al. Damascene benchmark of Ru, Co and Cu in scaled dimensions // 2018 IEEE International Interconnect Technology Conference (IITC). Piscataway, 2018: 172–174

[46] Milosevic E, Kerdsongpanya S, Zangiabadi A, et al. Resistivity size effect in epitaxial Ru (0001) layers. Journal of Applied Physics, 2018, 124(16): 165105

[47] Milosevic E, Kerdsongpanya S, McGahay M E, et al. Resistivity scaling and electron surface scattering in epitaxial Co (0001) layers. Journal of Applied Phy-sics, 2019, 125(24): 245105

Monte Carlo Approach for Nano-scale Metal Interconnect Simulation

HU Yuanzhao1, DU Gang2, YANG Yi1, ZHAO Yudi1,2, ZHAO Kai1,2,†

1. Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing Information Science and Technology University, Beijing 100101; 2. School of Electronic Engineering and Computer Science, Peking University, Beijing 100871; † Corresponding author, E-mail: zhaokai@pku.edu.cn

Abstract A Monte Carlo simulation method is proposed to simulate the electron transport characteristics in metal interconnects, with the presence of major scattering mechanisms. The results show that the plasma excimer scattering is the major contributor for bulk resistivity, followed by the electron-to-electron scattering. However, grain boundary scattering, dominating the resistivity of nanoscale interconnects, is the main reason for the “size effect”. The comparison of the simulated results and experimental data indicates that the proposed method can accurately calculate the resistivity for different metallic materials from bulk to nanoscale.

Key words Monte Carlo method; scattering; resistivity; interconnect

doi: 10.13209/j.0479-8023.2022.008

北京市教委科研计划(KM202111232016)资助

收稿日期: 2021-04-12;

修回日期: 2021-05-26