节点文献

室温下工作的非易失性分子级存储单元的操作设计与分子动力学模拟

Conceptual Design with a Molecular Dynamics Simulation of the Performance Characteristics of a Nonvolatile Molecular Memory Cell Operating in a Room-Temperature Environment

【作者】 叶森斌

【导师】 张文; 韩平畴;

【作者基本信息】 复旦大学 , 流体力学, 2009, 博士

【摘要】 半导体工业目前正面临着两大严峻的挑战:一方面是半导体集成电路的线宽尺寸将很快走向物理极限;另一方面是还没有可行性的半导体存储技术能够用于制造超高速的非易失性随机存储器(RAM)。在本文中,我们将提出两种碳纳米管分子级存储单元,来解决上述问题。碳纳米管分子级存储单元的读写操作主要通过脉冲电场(压)控制内管的往复振荡运动来实现。我们采用分子动力学模拟研究存储单元的工作性能,并发现我们提出的第二种存储单元,在室温下具有最可靠的超高速的非易失性工作特性。封闭型双层碳纳米管存储单元是我们提出的第一种非易失性分子级存储单元,由一个两端封闭的外管和一个两端封闭的内管组成。范德华势能的分子静力学计算结果表明,这种封闭双层碳纳米管具有双稳态结构,可以定义为“0”、“1”两个逻辑状态用于读写操作。逻辑状态问的开关电场要求达到1.5V/nm,用于驱动内管在状态点问的往复运动。为了实现非易失性的写入操作,本文提出了两种脉冲控制电场的加载方式,一种为初始激励型(short duration),另一种为过程保持型(10ng duration)。研究发现,尽管初始激励型控制方式具有较小的功耗,但对于外界的扰动影响比较敏感,而过程保持型具有比较可靠的写操作能力。为了研究封闭型存储单元在室温环境下的工作性能,本文分别用了Langevin、Gunsteren-Berendsen和Guo-Berendsen三种等温分子动力学方法对存储单元在室温下的工作特性进行模拟比较,结果发现,只有在Langevin等温模拟中,封闭型碳纳米管存储单元在室温下才表现出非易失性的存储功能,频率可以达到4GHz,而在其他两种方法的等温分子动力学模拟中,却没有实现高速非易失性存储的功能。本文提出的第二种设计,是由一个中间割断、两端开口的外管和一个两端封闭的内管以及两侧的读写电极组成,文中称为两段式开口型双层碳纳米管存储单元。把外管割成两段,是为了使存储单元具有比普通的开口型碳纳米管存储单元更深更宽的双稳态势阱;并且可以对两段外管问的间隔距离以及电极的沉积位置等结构尺寸参数进行优化,来改善存储单元的工作的稳定性和非易失性。根据对碳纳米管层问的范德华势能、碳管与金属电极问的结合能的叠加结果分析以及分子静力学计算发现,本文提出的两段式开口型存储单元具有较为宽深的双稳态势阱用于定义“0”、“1”状态,而且7.0V的开关电压可以充分驱动内管在状态点之间的运动。同样,我们研究了初始激励型(short duration)和过程保持型(10ngduration)两种电压脉冲控制方式来实现状态的准确转变。在Langevin分子动力学模拟中,两种控制方式对应的状态转变时间分别为50ps和40ps,我们采用过程保持型控制方式,由于它具有更可靠的写操作特性。与第一种设计的研究相同,我们也分别采用了另外两种等温控制方法来模拟两段式开口型双层碳纳米管存储单元的工作性能。结果发现,第二种存储单元设计能够在所有的三种等温分子动力学模拟中,在室温环境下均表现出稳定、可靠的非易失性存储功能,并且写操作频率可以高达16.6GHz,是目前主流的非易失性存储器闪存的100倍以上。因此,两段式开口型双层碳纳米管存储单元显然优于第一种设计的封闭型存储单元,更有希望发展成为未来的新型存储器,既可以作为大容量永久性存储设备,也可以应用到超高速的随机存储。另外,由于这种分子级存储单元是基于“Bottom—Up”的研究思路和制造技术,因此也可以解决目前传统半导体技术遇到集成芯片的线宽极限问题。

【Abstract】 The semiconductor industry is confronted with 2 significant challenges; the fast approaching miniaturization limits in memory technologies and the lack of a viable technology for introducing operational non-volatility in the random access memory (RAM) of a computer. In this dissertation we proposed 2 conceptual designs of a carbon nanotube (CNT)-based molecular memory cell, each of which is able to fully address both issues. The read-write operations of the memory cell are accomplished by the back-and-forth oscillations of the inner tube that is driven by a controlled electrostatic field impulse. To study its performance characteristics we employed molecular dynamics (MD) simulation that takes into account the main environmental forces such as those arising from the interactions between carbon-carbon atoms, and from the energies of the CNT-electrode binding and the external capacitive source. From the isothermal MD studies, we determined that our second design is most reliable in preserving the non-volatility at room temperatures.Our first design is a double-wall CNT-based memory cell made up of a fully-capped outer-tube and a fully-capped inner-tube or core. The molecular static calculations of the van der Waals potential indicate that the design is bistable with 2 well-defined logical states "0" and "1" for the read-write operations. The design requires an external electrostatic field of 1.5 V/nm to effect a switch between the 2 logical states. For the write operation, 2 types of external electrical field: a short-duration and a long-duration trapezoidal impulse were used. Although the former consumes less power, it is more sensitive to environmental disturbances. The latter on the other hand, uses more power but yields a better write reliability. To assess its performance at the room temperature 300K an isothermal MD simulation with the following 3 thermostat models applied in-turns: Langevin, Gunsteren-Berendsen and Guo-Berendsen was conducted. The Langevin MD study indicates that the fully-capped, double-wall CNT design behaves as a nonvolatile memory cell at room temperature with operating frequencies of up to 4 GHz. However, when replaced by either of the 2 remaining thermostat models, the fully-capped design is not only not able to function as a nonvolatile memory cell, it is also unable to perform the write operations at room temperature.The second design is an opened, center-split double-wall CNT-based memory cell with platinum electrodes placed at two extremities. The outer-tube is opened at both ends but the inner-tube remains fully-capped. To produce deeper and wider potential wells in the system energetics, we center-split the outer-tube into 2 sections. This design has the 2 structural parameters that need to be optimized: the split gap and the electrode gaps that determine the positioning of the 2 platinum electrodes. From a superposition of molecular static calculations of the van der Waals potential for each of the cell configuration, we showed that the design possesses bistable characteristics with 2 well-defined logical states "0" and "1". A switching voltage of 7.0 V is found to be sufficient to drive the core between the 2 logical states. Also, the same 2 external fields: a short-duration and a long-duration trapezoidal impulse were employed to realize the states transition. The Langevin MD results show that, the transition times of about 50ps and 40ps correspond to the short-duration and long-duration writing operations, respectively. Also, we adopted the long-duration impulse excitation in order to produce a more reliable write operation. As with the first design, an isothermal MD simulation at 300K with the same 3 thermostat models applied in-turns was conducted to assess the performance of the memory cell. The MD study showed that the opened, center-split memory cell produces stable, reliable and nonvolatile room-temperature operations for all 3 thermostat models. It is able to operate at frequencies of up to 16.6 GHz, which is at least 100 times faster than the flash memory. Therefore, the second design is clearly superior to the first proposal. We recommend the opened, center-split double-wall CNT memory cell as the more promising candidate for the further development of a novel memory device that can act both as a permanent terabit solid-state storage and a nonvolatile RAM. Additionally, it is molecular in scale based on the bottoms-up technology and hence, should not be handicapped by miniaturization limits in the foreseeable future.

  • 【网络出版投稿人】 复旦大学
  • 【网络出版年期】2010年 03期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络