资料同化中的数字滤波弱约束试验及分析

气象数值预报中,由于分析过程引入初始非平衡,从而引起虚假快波振荡,重力波控制弱约束把资料分析过程和初始化过程结合在一起,通过数字滤波弱约束在极小化过程中实现对分析场的平衡约束,克服非平衡问题. 以2008年初的一次南方雨雪天气为研究个例,进行了数字滤波弱约束的同化试验和预报试验,结果表明,数字滤波弱约束4D-Var能充分控制快波振荡的出现和初始调整现象,使得到的分析场不仅能更好的逼近观测,而且能更好地与模式动力相协调. 预报检验的结果表明,在同化过程中施加数字滤波弱约束,能有效滤除由于地形或观测资料等因素 关键词： 变分同化 初始非平衡 数字滤波 弱约束     

Crystallization of Ge2Sb2Te5 phase—change optical disk media

In this paper, the crystallization behaviour of amorphous Ge₂Sb₂Te₅ thin films is investigated using differential scanning calorimetry), x-ray diffraction and optical transmissivity measurements. It is indicated that only the amorphous phase to face-centred-cubic phase transformation occurs during laser annealing of the normal phase-change structure, which is a benefit for raising the phase-change optical disk's carrier-to-noise ratio (CNR). For amorphous Ge₂Sb₂Te₅ thin films, the crystallization temperature is about 200℃ and the melting temperature is 546.87℃. The activation energy for the crystallization, E_a, is 2.25eV. The crystallization dynamics for Ge₂Sb₂Te₅ thin films obeys the law of nucleation and growth reaction. The sputtered Ge₂Sb₂Te₅ films were initialized by an initializer unit. The initialization conditions have a great effect on the reflectivity contrast of the Ge₂Sb₂Te₅ phase-change optical disk.     

Consistent Initialization for Nonlinear Index-2 Differential–Algebraic Equation: Large Sparse Systems in MATLAB

An important component of any initial-value solver for higher-index differential–algebraic equations consists in the computation of consistent initial values. In a recent paper [5], an algorithm is proposed which is applicable to a very general class of index-2 systems. Unfortunately, the computational expense is rather high. We present a modification of this approach, which gives rise to a MATLAB implementation capable of handling systems of moderate dimension (several thousands of unknowns). The algorithm is illustrated by examples.     

光学双稳与混沌实验

借助自主研制的"液晶光电混合型光学双稳与混沌系统"研究了光学双稳态与混沌运动的一般规律.在有反馈的条件下,对于1个入射光强,存在着2个不同的透射光强,并以滞后迴线形式为特征,即系统呈现双稳态.依然在有反馈的条件下,系统的运动状态随着延迟时间的增加,由周期振荡、倍周期分岔演化到混沌,而且,混沌对于初始值具有高度敏感性.     

关于量子体系的高效相干能量转移的最优初始化的研究

For an energy transfer network, the irreversible depletion of excited electron energy occurs through either an efficient flow into an outer energy sink or an inefficient decay. With a small decay rate, the energy transfer efficiency is quantitatively reflected by the average life time of excitation energy before being trapped in the sink where the decay process is omitted. In the weak dissipation regime, the trapping time is analyzed within the exciton population subspace based on the secular Redfield equation. The requirement of the noise-enhanced energy transfer is obtained, where the trapping time follows an exact or approximate 1/Γ-scaling of the dissipation strength Γ. On the opposite side, optimal initial system states are conceptually constructed to suppress the 1/Γ-scaling of the trapping time and maximize the coherent transfer efficiency. Our theory is numerically testified in four models, including a biased two-site system, a symmetric three-site branching system, a homogeneous onedimensional chain, and an 8-chromophore FMO protein complex.     

Optimal Initialization of a Quantum System for an Efficient Coherent Energy Transfer?

For an energy transfer network, the irreversible depletion of excited electron energy occurs through either an efficient flow into an outer energy sink or an inefficient decay. With a small decay rate, the energy transfer efficiency is quantitatively reflected by the average life time of excitation energy before being trapped in the sink where the decay process is omitted. In the weak dissipation regime, the trapping time is analyzed within the exciton population subspace based on the secular Redfield equation. The requirement of the noise-enhanced energy transfer is obtained, where the trapping time follows an exact or approximate 1/Γ-scaling of the dissipation strength Γ. On the opposite side, optimal initial system states are conceptually constructed to suppress the 1/Γ-scaling of the trapping time and maximize the coherent transfer efficiency. Our theory is numerically testified in four models, including a biased two-site system, a symmetric three-site branching system, a homogeneous onedimensional chain, and an 8-chromophore FMO protein complex.     

A stochastic model of deposition processes with nucleation

We study an interacting particle system on a one-dimensional infinite lattice and one-dimensional lattices with a periodic boundary. In this system, each site of the lattice may be either empty or occupied and initially all the lattice sites are empty. The evolution of the system is defined as follows: an empty site waits an exponential time with mean 1 and becomes occupied, and an occupied site becomes empty at a time which is distributed exponentially with mean _k, wherek is the number of occupied neighboring sites of this site in the current state of the system. We show that the mean number of the occupied sites of the lattice, considered as a function of time, may possess a convex part. A sufficient condition for this is that ₀ is large and _k,k1, are small. The studied system has been proposed recently as a mathematical model of certain deposition processes, in particular those which exhibit nucleation caused by lateral attractive interaction between the deposited molecules. Our research was motivated by the observation that the density of deposited molecules contains a convex part, over some time interval, if the attractive forces are strong, while this density is a concave function of time if these forces are weak or absent. Our result agrees with this observation.     

k-均值算法的初始化方法综述

k-均值问题自提出以来一直吸引组合优化和计算机科学领域的广泛关注, 是经典的NP-难问题之一. 给定N个d维实向量构成的观测集, 目标是把这N个观测点划分到k(\leq N)个集合中, 使得所有集合中的点到对应的聚类中心距离的平方和最小, 一个集合的聚类中心指的是该集合 中所有观测点的均值. k-均值算法作为解决k-均值问题的启发式算法,在实际应用中因其出色的收敛速度而倍受欢迎. k-均值算法可描述为: 给定问题的初始化分组, 交替进行指派(将观测点分配到离其最近的均值点)和更新(计算新的聚类的均值点)直到收敛到某一解. 该算法通常被认为几乎是线性收敛的. 但缺点也很明显, 无法保证得到的是全局最优解, 并且算法结果好坏过于依赖初始解的选取. 于是学者们纷纷提出不同的初始化方法来提高k-均值算法的质量. 现筛选和罗列了关于选取初始解的k-均值算法的初始化方法供读者参考.     

A corrected symmetric SPH method to simulate viscoelastic free surface flows based on the PTT model

In this work, a corrected symmetric and periodic density reinitialized SPH (CSPDR‐SPH) method is proposed and extended to simulate the viscoelastic free surface flows based on the Phan–Thien–Tanner model. The improvements mainly lie in deriving a corrected symmetric kernel gradient, and combining it with a periodic density reinitialization procedure. In addition, a simple artificial viscosity and a simple artificial stress form are adopted. Thus, the CSPDR‐SPH method has higher accuracy and better stability than the SPH method, and conserves both linear and angular momentums. The consistency and convergence of the CSPDR‐SPH method are justified by approximating a function in one and two dimensions. The merits of CSPDR‐SPH method are demonstrated by several benchmarks. The simple flow in a two‐dimensional channel is investigated to show the capability of the CSPDR‐SPH method to simulate the viscoelastic free surface flow. Then the CSPDR‐SPH method is extended to simulate the impacting drop problem. Numerical results show that the CSPDR‐SPH method can precisely capture the viscoelastic free surface. The Reynolds number, Weissenberg number and elongation parameter have remarkable effect on the flows. Copyright © 2012 John Wiley & Sons, Ltd.     

Initialization behavior at various target molecular weight RAFT‐mediated methyl acrylate polymerizations

In the reversible addition‐fragmentation chain transfer (RAFT) mediated polymerization of methyl acrylate, a selective reaction is observed in the early stages of the polymerization. This initialization process was earlier observed in in situ ¹H NMR spectroscopy experiments where extremely low target molar masses were chosen (around DP = 5). Here, for the first time, the presence of the initialization process is identified as the cause of an induction period under typical conditions of a RAFT‐mediated polymerization. © 2008 Wiley Periodicals, Inc. J Polym Sci Part A: Polym Chem 46: 2500–2509, 2008