A fault-tolerant technique for nanocomputers Nand. Nanotechnology

合集下载

一种低K工艺产品分层问题的解决方案

一种低K工艺产品分层问题的解决方案杜新，王西国（北京中电华大电子设计有限责任公司，北京，100102）（射频识别芯片检测技术北京市重点实验室）摘要：随着半导体工艺的不断进步，芯片线宽越来越小，RC延迟成为制约集成电路性能进一步提高的关键性因素。

低K工艺技术的出现虽然解决了RC延迟问题，但它为后续的封装工艺却带来以下问题：即由于材料本身的特性（与金属层较弱的粘结力，较弱的机械强度），如果采用金刚石双刀切割工艺，很容易在金属层与内层介电层（Inner layer dielectric，ILD）层之间产生分层或剥离的现象。

所以，探索新的晶圆切割方法对低K工艺技术的引入显得越来越重要。

本文介绍一种低K工艺产品分层问题的解决方案。

关键词：低K工艺；晶圆切割；分层；激光减划A solution to the stratification problem of low K process productsDU Xin,WANG Xi-guo（CEC Huada Electronic Design Co.,Ltd.,Beijing100102,China）（Beijing Key Laboratory of RFID Chip Test Technology）Abstract:With the continuous progress of semiconductor technology,the line width of chip is getting smaller and smaller.RC delay has become the key factor that restricts the further improvement of the performance of integrated circuits.Although the emergence of low K process technology solves the problem of RC delay,it brings the following problems to the subsequent packaging process:it is due to the characteristics of the material itself（weak adhesion with metal layer,weak mechanical strength）,if diamond double knife cutting process is used,it is easy to delaminate or peel between metal layer and ILD layer.It is more and more important to explore new wafer cutting method for the introduction of low K process technology.This paper introduces a solution to the stratification problem of low K pro-cess produces.Key words:Low K process;wafer cutting;delamination;laser reduction1引言随着半导体技术的不断进步，低K 工艺被广泛应用到集成电路中，而该工艺的Low-k 材料研发本身的挑战在于：在获得所需要的低介电常数的同时，Low-k 材料还必须满足良好的热和机械特性。

微型生物传感器(英文)

Applications of Nanobiosensors
Biological Applications DNA Sensors; Genetic monitoring, disease Immunosensors; HIV, Hepatitis,other viral diseas, drug testing, environmental monitoring… Cell-based Sensors; functional sensors, drug testing… Point-of-care sensors; blood, urine, electrolytes, gases, steroids, drugs, hormones, proteins, other… Bacteria Sensors; (E-coli, streptococcus, other): food industry, medicine, environmental, other. Enzyme sensors; diabetics, drug testing, other. Environmental Applications Detection of environmental pollution and toxicity Agricultural monitoring Ground water screening Ocean monitoring
Theory
Principle of Detection Piezoelectric Mass Electrochemical Electric distribution Optical Light intensity Calorimetric Heat
Types of Nanobiosensors Optical Biosensors Nanotube Based Biosensors Electrical Biosensors Viral Nanosensors Electrochemical Biosensors Nanoshell Biosensors Nanowire Biosensors

Practical byzantine fault tolerance

Appears in the Proceedings of the Third Symposium on Operating Systems Design and Implementation,New Orleans,USA,February1999Practical Byzantine Fault ToleranceMiguel Castro and Barbara LiskovLaboratory for Computer Science,Massachusetts Institute of Technology,545Technology Square,Cambridge,MA02139castro,liskov@AbstractThis paper describes a new replication algorithm that is able to tolerate Byzantine faults.We believe that Byzantine-fault-tolerant algorithms will be increasingly important in the future because malicious attacks and software errors are increasingly common and can cause faulty nodes to exhibit arbitrary behavior.Whereas previous algorithms assumed a synchronous system or were too slow to be used in practice, the algorithm described in this paper is practical:it works in asynchronous environments like the Internet and incorporates several important optimizations that improve the response time of previous algorithms by more than an order of magnitude.We implemented a Byzantine-fault-tolerant NFS service using our algorithm and measured its performance.The results show that our service is only3%slower than a standard unreplicated NFS. 1IntroductionMalicious attacks and software errors are increasingly common.The growing reliance of industry and gov-ernment on online information services makes malicious attacks more attractive and makes the consequences of successful attacks more serious.In addition,the number of software errors is increasing due to the growth in size and complexity of software.Since malicious attacks and software errors can cause faulty nodes to exhibit Byzan-tine(i.e.,arbitrary)behavior,Byzantine-fault-tolerant al-gorithms are increasingly important.This paper presents a new,practical algorithm for state machine replication[17,34]that tolerates Byzantine faults.The algorithm offers both liveness and safety provided at most1This research was supported in part by DARPA under contract DABT63-95-C-005,monitored by Army Fort Huachuca,and under contract F30602-98-1-0237,monitored by the Air Force Research Laboratory, and in part by NEC.Miguel Castro was partially supported by a PRAXIS XXI fellowship.and replication techniques that tolerate Byzantine faults (starting with[19]).However,most earlier work(e.g., [3,24,10])either concerns techniques designed to demonstrate theoretical feasibility that are too inefﬁcient to be used in practice,or assumes synchrony,i.e., relies on known bounds on message delays and process speeds.The systems closest to ours,Rampart[30]and SecureRing[16],were designed to be practical,but they rely on the synchrony assumption for correctness,which is dangerous in the presence of malicious attacks.An attacker may compromise the safety of a service by delaying non-faulty nodes or the communication between them until they are tagged as faulty and excluded from the replica group.Such a denial-of-service attack is generally easier than gaining control over a non-faulty node.Our algorithm is not vulnerable to this type of attack because it does not rely on synchrony for safety.In addition,it improves the performance of Rampart and SecureRing by more than an order of magnitude as explained in Section7.It uses only one message round trip to execute read-only operations and two to execute read-write operations.Also,it uses an efﬁcient authentication scheme based on message authentication codes during normal operation;public-key cryptography,which was cited as the major latency[29] and throughput[22]bottleneck in Rampart,is used only when there are faults.To evaluate our approach,we implemented a replica-tion library and used it to implement a real service:a Byzantine-fault-tolerant distributedﬁle system that sup-ports the NFS protocol.We used the Andrew bench-mark[15]to evaluate the performance of our system.The results show that our system is only3%slower than the standard NFS daemon in the Digital Unix kernel during normal-case operation.Thus,the paper makes the following contributions:It describes theﬁrst state-machine replication proto-col that correctly survives Byzantine faults in asyn-chronous networks.It describes a number of important optimizations that allow the algorithm to perform well so that it can be used in real systems.It describes the implementation of a Byzantine-fault-tolerant distributedﬁle system.It provides experimental results that quantify the cost of the replication technique.The remainder of the paper is organized as follows. We begin by describing our system model,including our failure assumptions.Section3describes the problem solved by the algorithm and states correctness conditions. The algorithm is described in Section4and some important optimizations are described in Section 5. Section6describes our replication library and how we used it to implement a Byzantine-fault-tolerant NFS.Section7presents the results of our experiments. Section8discusses related work.We conclude with a summary of what we have accomplished and a discussion of future research directions.2System ModelWe assume an asynchronous distributed system where nodes are connected by a network.The network may fail to deliver messages,delay them,duplicate them,or deliver them out of order.We use a Byzantine failure model,i.e.,faulty nodes may behave arbitrarily,subject only to the restriction mentioned below.We assume independent node failures. For this assumption to be true in the presence of malicious attacks,some steps need to be taken,e.g.,each node should run different implementations of the service code and operating system and should have a different root password and a different administrator.It is possible to obtain different implementations from the same code base[28]and for low degrees of replication one can buy operating systems from different vendors.N-version programming,i.e.,different teams of programmers produce different implementations,is another option for some services.We use cryptographic techniques to prevent spooﬁng and replays and to detect corrupted messages.Our messages contain public-key signatures[33],message authentication codes[36],and message digests produced by collision-resistant hash functions[32].We denote a message signed by node as and the digest of message by.We follow the common practice of signing a digest of a message and appending it to the plaintext of the message rather than signing the full message(should be interpreted in this way).All replicas know the others’public keys to verify signatures. We allow for a very strong adversary that can coordinate faulty nodes,delay communication,or delay correct nodes in order to cause the most damage to the replicated service.We do assume that the adversary cannot delay correct nodes indeﬁnitely.We also assume that the adversary(and the faulty nodes it controls)are computationally bound so that(with very high probability)it is unable to subvert the cryptographic techniques mentioned above.For example,the adversary cannot produce a valid signature of a non-faulty node, compute the information summarized by a digest from the digest,orﬁnd two messages with the same digest. The cryptographic techniques we use are thought to have these properties[33,36,32].3Service PropertiesOur algorithm can be used to implement any deterministic replicated service with a state and some operations.The operations are not restricted to simple reads or writes of portions of the service state;they can perform arbitrary deterministic computations using the state and operation arguments.Clients issue requests to the replicated service to invoke operations and block waiting for a reply.The replicated service is implemented by replicas.Clients and replicas are non-faulty if they follow the algorithm in Section4and if no attacker can forge their signature. The algorithm provides both safety and liveness assum-ing no more than13replicas are faulty and delay does notgrow faster than indeﬁnitely.Here,delay is the time between the moment when a message is sent fortheﬁrst time and the moment when it is received by its destination(assuming the sender keeps retransmitting themessage until it is received).(A more precise deﬁnition can be found in[4].)This is a rather weak synchronyassumption that is likely to be true in any real system provided network faults are eventually repaired,yet it enables us to circumvent the impossibility result in[9].The resiliency of our algorithm is optimal:31is the minimum number of replicas that allow an asynchronoussystem to provide the safety and liveness properties when up to replicas are faulty(see[2]for a proof).Thismany replicas are needed because it must be possible to proceed after communicating with replicas,since replicas might be faulty and not responding.However, it is possible that the replicas that did not respond arenot faulty and,therefore,of those that responded might be faulty.Even so,there must still be enough responsesthat those from non-faulty replicas outnumber those from faulty ones,i.e.,2.Therefore3.The algorithm does not address the problem of fault-tolerant privacy:a faulty replica may leak information to an attacker.It is not feasible to offer fault-tolerant privacy in the general case because service operations may perform arbitrary computations using their arguments and the service state;replicas need this information in the clear to execute such operations efﬁciently.It is possible to use secret sharing schemes[35]to obtain privacy even in the presence of a threshold of malicious replicas[13] for the arguments and portions of the state that are opaque to the service operations.We plan to investigate these techniques in the future.4The AlgorithmOur algorithm is a form of state machine replication[17,34]:the service is modeled as a state machine that is replicated across different nodes in a distributed system. Each state machine replica maintains the service state and implements the service operations.We denote the set of replicas by and identify each replica using an integer in01.For simplicity,we assume 31where is the maximum number of replicas that may be faulty;although there could be more than31replicas,the additional replicas degrade performance(since more and bigger messages are being exchanged)without providing improved resiliency.The replicas move through a succession of conﬁgura-tions called views.In a view one replica is the primary and the others are backups.Views are numbered con-secutively.The primary of a view is replica such that mod,where is the view number.View changes are carried out when it appears that the primary has failed.Viewstamped Replication[26]and Paxos[18]used a similar approach to tolerate benign faults(as dis-cussed in Section8.)The algorithm works roughly as follows:1.A client sends a request to invoke a service operationto the primary2.The primary multicasts the request to the backups3.Replicas execute the request and send a reply to theclient4.The client waits for1replies from differentreplicas with the same result;this is the result of the operation.Like all state machine replication techniques[34], we impose two requirements on replicas:they must be deterministic(i.e.,the execution of an operation in a given state and with a given set of arguments must always produce the same result)and they must start in the same state.Given these two requirements,the algorithm ensures the safety property by guaranteeing that all non-faulty replicas agree on a total order for the execution of requests despite failures.The remainder of this section describes a simpliﬁed version of the algorithm.We omit discussion of how nodes recover from faults due to lack of space.We also omit details related to message retransmissions. Furthermore,we assume that message authentication is achieved using digital signatures rather than the more efﬁcient scheme based on message authentication codes; Section5discusses this issue further.A detailed formalization of the algorithm using the I/O automaton model[21]is presented in[4].4.1The ClientA client requests the execution of state machine operation by sending a REQUEST message to the primary.Timestamp is used to ensure exactly-once semantics for the execution of client requests. Timestamps for’s requests are totally ordered such that later requests have higher timestamps than earlier ones; for example,the timestamp could be the value of the client’s local clock when the request is issued.Each message sent by the replicas to the client includes the current view number,allowing the client to track the view and hence the current primary.A client sends a request to what it believes is the current primary using a point-to-point message.The primary atomically multicasts the request to all the backups using the protocol described in the next section.A replica sends the reply to the request directly to the client.The reply has the form REPLYwhere is the current view number,is the timestamp of the corresponding request,is the replica number,and is the result of executing the requested operation.The client waits for1replies with valid signatures from different replicas,and with the same and,beforeaccepting the result.This ensures that the result is valid, since at most replicas can be faulty.If the client does not receive replies soon enough,it broadcasts the request to all replicas.If the request has already been processed,the replicas simply re-send the reply;replicas remember the last reply message they sent to each client.Otherwise,if the replica is not the primary, it relays the request to the primary.If the primary does not multicast the request to the group,it will eventually be suspected to be faulty by enough replicas to cause a view change.In this paper we assume that the client waits for one request to complete before sending the next one.But we can allow a client to make asynchronous requests,yet preserve ordering constraints on them.4.2Normal-Case OperationThe state of each replica includes the state of the service,a message log containing messages the replica has accepted,and an integer denoting the replica’s current view.We describe how to truncate the log in Section4.3.When the primary,,receives a client request,, it starts a three-phase protocol to atomically multicast the request to the replicas.The primary starts the protocol immediately unless the number of messages for which the protocol is in progress exceeds a given maximum.In this case,it buffers the request.Buffered requests are multicast later as a group to cut down on message trafﬁc and CPU overheads under heavy load;this optimization is similar to a group commit in transactional systems[11].For simplicity,we ignore this optimization in the description below.The three phases are pre-prepare,prepare,and commit. The pre-prepare and prepare phases are used to totally order requests sent in the same view even when the primary,which proposes the ordering of requests,is faulty.The prepare and commit phases are used to ensure that requests that commit are totally ordered across views.In the pre-prepare phase,the primary assigns a sequence number,,to the request,multicasts a pre-prepare message with piggybacked to all the backups, and appends the message to its log.The message has the form PRE-PREPARE,where indicates the view in which the message is being sent,is the client’s request message,and is’s digest.Requests are not included in pre-prepare messages to keep them small.This is important because pre-prepare messages are used as a proof that the request was assigned sequence number in view in view changes. Additionally,it decouples the protocol to totally order requests from the protocol to transmit the request to the replicas;allowing us to use a transport optimized for small messages for protocol messages and a transport optimized for large messages for large requests.A backup accepts a pre-prepare message provided:the signatures in the request and the pre-prepare message are correct and is the digest for;it is in view;it has not accepted a pre-prepare message for viewand sequence number containing a different digest;the sequence number in the pre-prepare message is between a low water mark,,and a high water mark,.The last condition prevents a faulty primary from exhausting the space of sequence numbers by selecting a very large one.We discuss how and advance in Section4.3.If backup accepts the PRE-PREPARE message,it enters the prepare phase by multicasting a PREPARE message to all other replicas and adds both messages to its log.Otherwise,it does nothing.A replica(including the primary)accepts prepare messages and adds them to its log provided their signatures are correct,their view number equals the replica’s current view,and their sequence number is between and.We deﬁne the predicate prepared to be true if and only if replica has inserted in its log:the request ,a pre-prepare for in view with sequence number ,and2prepares from different backups that match the pre-prepare.The replicas verify whether the prepares match the pre-prepare by checking that they have the same view,sequence number,and digest.The pre-prepare and prepare phases of the algorithm guarantee that non-faulty replicas agree on a total order for the requests within a view.More precisely,they ensure the following invariant:if prepared is true then prepared is false for any non-faulty replica(including)and any such that.This is true because prepared and31imply that at least1non-faulty replicas have sent a pre-prepare or prepare for in view with sequence number.Thus,for preparedto be true at least one of these replicas needs to have sent two conﬂicting prepares(or pre-prepares if it is the primary for),i.e.,two prepares with the same view and sequence number and a different digest.But this is not possible because the replica is not faulty.Finally,our assumption about the strength of message digests ensures that the probability that and is negligible.Replica multicasts a COMMIT to the other replicas when prepared becomes true. This starts the commit phase.Replicas accept commit messages and insert them in their log provided they are properly signed,the view number in the message is equal to the replica’s current view,and the sequence number is between andWe deﬁne the committed and committed-local predi-cates as follows:committed is true if and only if prepared is true for all in some set of 1non-faulty replicas;and committed-localis true if and only if prepared is true and has accepted21commits(possibly including its own) from different replicas that match the pre-prepare for;a commit matches a pre-prepare if they have the same view,sequence number,and digest.The commit phase ensures the following invariant:if committed-local is true for some non-faulty then committed is true.This invariant and the view-change protocol described in Section4.4ensure that non-faulty replicas agree on the sequence numbers of requests that commit locally even if they commit in different views at each replica.Furthermore,it ensures that any request that commits locally at a non-faulty replica will commit at1or more non-faulty replicas eventually.Each replica executes the operation requested by after committed-local is true and’s state reﬂects the sequential execution of all requests with lower sequence numbers.This ensures that all non-faulty replicas execute requests in the same order as required to provide the safety property.After executing the requested operation,replicas send a reply to the client. Replicas discard requests whose timestamp is lower than the timestamp in the last reply they sent to the client to guarantee exactly-once semantics.We do not rely on ordered message delivery,and therefore it is possible for a replica to commit requests out of order.This does not matter since it keeps the pre-prepare,prepare,and commit messages logged until the corresponding request can be executed.Figure1shows the operation of the algorithm in the normal case of no primary faults.Replica0is the primary, replica3is faulty,and is the client.C123Figure1:Normal Case Operation4.3Garbage CollectionThis section discusses the mechanism used to discard messages from the log.For the safety condition to hold, messages must be kept in a replica’s log until it knows that the requests they concern have been executed by at least 1non-faulty replicas and it can prove this to others in view changes.In addition,if some replica misses messages that were discarded by all non-faulty replicas, it will need to be brought up to date by transferring all or a portion of the service state.Therefore,replicas also need some proof that the state is correct.Generating these proofs after executing every opera-tion would be expensive.Instead,they are generated periodically,when a request with a sequence number di-visible by some constant(e.g.,100)is executed.We will refer to the states produced by the execution of these re-quests as checkpoints and we will say that a checkpoint with a proof is a stable checkpoint.A replica maintains several logical copies of the service state:the last stable checkpoint,zero or more checkpoints that are not stable,and a current state.Copy-on-write techniques can be used to reduce the space overhead to store the extra copies of the state,as discussed in Section6.3.The proof of correctness for a checkpoint is generated as follows.When a replica produces a checkpoint, it multicasts a message CHECKPOINT to the other replicas,where is the sequence number of the last request whose execution is reﬂected in the state and is the digest of the state.Each replica collects checkpoint messages in its log until it has21of them for sequence number with the same digest signed by different replicas(including possibly its own such message).These21messages are the proof of correctness for the checkpoint.A checkpoint with a proof becomes stable and the replica discards all pre-prepare,prepare,and commit messages with sequence number less than or equal to from its log;it also discards all earlier checkpoints and checkpoint messages.Computing the proofs is efﬁcient because the digest can be computed using incremental cryptography[1]as discussed in Section6.3,and proofs are generated rarely. The checkpoint protocol is used to advance the low and high water marks(which limit what messages will be accepted).The low-water mark is equal to the sequence number of the last stable checkpoint.The high water mark,where is big enough so that replicas do not stall waiting for a checkpoint to become stable.For example,if checkpoints are taken every100 requests,might be200.4.4View ChangesThe view-change protocol provides liveness by allowing the system to make progress when the primary fails.View changes are triggered by timeouts that prevent backups from waiting indeﬁnitely for requests to execute.A backup is waiting for a request if it received a valid requestand has not executed it.A backup starts a timer when itreceives a request and the timer is not already running. It stops the timer when it is no longer waiting to executethe request,but restarts it if at that point it is waiting to execute some other request.If the timer of backup expires in view,the backup starts a view change to move the system toview 1.It stops accepting messages(other thancheckpoint,view-change,and new-view messages)and multicasts a VIEW-CHANGE1messageto all replicas.Here is the sequence number of the last stable checkpoint known to,is a set of21validcheckpoint messages proving the correctness of,and is a set containing a set for each request that prepared at with a sequence number higher than.Each set contains a valid pre-prepare message(without thecorresponding client message)and2matching,valid prepare messages signed by different backups with the same view,sequence number,and the digest of. When the primary of view1receives2valid view-change messages for view1from other replicas, it multicasts a NEW-VIEW1message to all other replicas,where is a set containing the valid view-change messages received by the primary plus the view-change message for1the primary sent(or would have sent),and is a set of pre-prepare messages(without the piggybacked request).is computed as follows:1.The primary determines the sequence number min-sof the latest stable checkpoint in and the highestsequence number max-s in a prepare message in.2.The primary creates a new pre-prepare message forview1for each sequence number between min-s and max-s.There are two cases:(1)there is at least one set in the component of some view-change message in with sequence number,or(2)there is no such set.In theﬁrst case,the primary createsa new message PRE-PREPARE1,whereis the request digest in the pre-prepare message for sequence number with the highest view number in.In the second case,it creates a new pre-prepare message PRE-PREPARE1, where is the digest of a special null request;a null request goes through the protocol like otherrequests,but its execution is a no-op.(Paxos[18] used a similar technique toﬁll in gaps.)Next the primary appends the messages in to itslog.If min-s is greater than the sequence number of its latest stable checkpoint,the primary also inserts the proof of stability for the checkpoint with sequence number min-s in its log,and discards information from the log as discussed in Section4.3.Then it enters view1:at this point it is able to accept messages for view 1.A backup accepts a new-view message for view1 if it is signed properly,if the view-change messages it contains are valid for view1,and if the set is correct;it veriﬁes the correctness of by performing acomputation similar to the one used by the primary to create.Then it adds the new information to its log asdescribed for the primary,multicasts a prepare for each message in to all the other replicas,adds these preparesto its log,and enters view 1.Thereafter,the protocol proceeds as described inSection4.2.Replicas redo the protocol for messages between min-s and max-s but they avoid re-executingclient requests(by using their stored information about the last reply sent to each client).A replica may be missing some request messageor a stable checkpoint(since these are not sent in new-view messages.)It can obtain missing information from another replica.For example,replica can obtain amissing checkpoint state from one of the replicas whose checkpoint messages certiﬁed its correctness in .Since1of those replicas are correct,replica will always obtain or a later certiﬁed stable checkpoint.Wecan avoid sending the entire checkpoint by partitioning the state and stamping each partition with the sequence number of the last request that modiﬁed it.To bring a replica up to date,it is only necessary to send it the partitions where it is out of date,rather than the whole checkpoint.4.5CorrectnessThis section sketches the proof that the algorithmprovides safety and liveness;details can be found in[4].4.5.1SafetyAs discussed earlier,the algorithm provides safety if all non-faulty replicas agree on the sequence numbers of requests that commit locally.In Section4.2,we showed that if preparedis true,prepared is false for any non-faulty replica(including)and any such that.This implies that two non-faulty replicas agree on the sequence number of requests that commit locally in the same view at the two replicas. The view-change protocol ensures that non-faulty replicas also agree on the sequence number of requests that commit locally in different views at different replicas.A request commits locally at a non-faulty replica withsequence number in view only if committedis true.This means that there is a set1containing at least 1non-faulty replicas such that preparedis true for every replica in the set.Non-faulty replicas will not accept a pre-prepare for view without having received a new-view message for(since only at that point do they enter the view).But any correct new-view message for view contains correct view-change messages from every replica in a。

ZnO Nanowire UV Photodetectors with High Internal Gain(中文翻译)

图 3.
(a)单根 ZnO NW 器件在 5V 偏压下光电流作为激发强度函数 (b)在 ZnO NW 中估计的光电导增益与光子吸收率之比
为了确定载流子寿命 T1，我们在低的激发强度下使用时间分辨的设备研究了光电流驰豫。图 4a 显示了不同片压下光电流在持续光照下（λ=390nm）的上升和入射光管断后光电流的衰减。通过对比与不同偏压下的曲线，发现光电流动力学在整个所研究的|V|<5V 偏压范围内独立于信号和外部电场强度，表明不存在空间电荷效应。通过双指数上升和下降函数获取的实线显示与数据符合的最好，我们推导出的加权平均光电流上升和下降时间常数是 τrise=23s 和 τdecay=33s。从常规的光电探测器 3dB 带宽表达式 B=1/2πT1，以及载流子寿命实验值（T1=33s），我们 X 获得了 ZnO NW 的 B≈5X10-3HZ。增益带宽积通过
N
F
AL
Hale Waihona Puke T1 ( F )T1 ( F ) T10
1 1 ( F / F0 )n
（1 ）
此处 Ƞ 是载流子光生量子效率，F 是光吸收率，A 和 L 分别是 NW 横截面和长度，T1 是载流子寿命。当陷阱充满后，能带变平，自由空穴数目增多；这增加了电子-空穴复合的几率，（which is manifested in an absorption-rate-dependent carrier lifetime）在吸附率依赖的载流子寿命 T1 (F)中很明显。T1 (F)的解析表达式在式(1)中给出，式中 T10 是在低激发密度下载流子寿命，F0 是当陷阱达到饱和时光子吸收率，n 是现象调整参数。从一般的光电流（Iph）表达式我们可以得到：
亚毫秒范围内：比如，在 ν=3KHz 是，G～2X106，使得 GB=6G Hz。从式 4，可推导出对极间距离为 2μm，电压 V=5V 时，载流子渡越时间为 Tt～30ps，对应载流子速度为 μ～ 270cm2/V· s。因为这些 NW 器件在 V=5V 时的暗电流接近 10nA（图 1），我们可以估计本 13 -3 征载流子浓度低至 n～10 cm ，意味着在本研究中所用的 ZnO NW 具有很高的结晶质量，并且解释了观察到非常高的光电导增益。

各种纳米结构

Less-Common Nanostructures in the Forms of VegetationOxana V.Kharissova and Boris I.Kharisov*Uni V ersidad Auto ´noma de Nue V o Leo ´n,Ciudad Uni V ersitaria UANL,San Nicola ´s de los Garza,N.L.66450MexicoNanostructures in the forms of varieties of vegetation are reviewed,in particular nanotrees and their agglomerates,nanobushes,nanomushrooms,nanoﬂowers,nanograss,and nanosheafs,among many others.Synthesis methods,inﬂuence of reaction conditions on the structural types of products,properties,formation mechanisms,and current and future applications are discussed.These nanostructures possess a series of useful properties (magnetic,semiconducting,ﬁeld-emission,etc.)and applications,mainly in catalysis,as well as for creation of nanomaterials,solar cells,and in medicine for tumor treatments,among other uses.IntroductionWhen the nanotechnology area began to develop intensively as an independent ﬁeld in the frontiers of physics,chemistry,materials chemistry and physics,medicine,biology,and other disciplines two decades ago,such terms as “nanoparticle”,“nanopowder”,“nanotube”,“nanowire”,“nanoring”,“nanorod”,“nanobelt”,“nanocluster”,“nanosheet”,“nanodot”,“nanoalloy”,“nanoribbon”,“nanoplate”,“nanoneedle”,“nanocone”,“nanoﬁl-ament”,“nanowall”,and “nanoﬁlm”rapidly become very common,especially the ﬁrst four nanoforms.For instance,a simple search using SciFinder results in hundreds of thousands of articles with keywords “nanoparticle”or “nanotube”.At the same time,during the last years,efforts of researchers have led to reports of enormous number of the nanostructure types above and the discovery of more rare species,such as “nanodumb-bells”,“nanoﬂowers”,“nanorices”,“nanolines”,“nanotowers”,“nanoshuttles”,“nanobowlings”,“nanowheels”,“nanofans”,“nanopencils”,“nanotrees”,“nanoarrows”,“nanonails”,“nano-bottles”,or “nanovolcanoes”,among many others.The problem of how to name a discovered rare nanostructure is commonly resolved according to the imagination of researchers.Because any novel nanoform/nanostructure could theoretically become useful,unexpected and unpredictable applications (as for example famous graphene,discovered not long ago 1),each new achievement is welcome due to an extreme importance of nanotechnology at this moment and in the future.Without a good understanding of the reasons for shape formation,ap-proaches to the synthesis of nanostructures can be hard to carry out.According to the available literature,no universal generaliza-tion of rare and common nanostructures is currently observed.Several existing classiﬁcations are related to dimensionality of the nanostructure itself and its components 2,3(for instance,0D clusters and particles,1D nanotubes and nanowires,2D nano-plates and layers,3D core/shell nanoparticles or self-assembled massives,intermediate dimensional nanostructures as fractals or dendrimers)or the classiﬁcation based on the triad symmetry group -shell composition -structural formula of the shell (here nanostructures are divided in branches,classes,and subclasses determined by the symmetry group of a shell and the sets of the quantum numbers of a structure).4In addition,a nonformal classiﬁcation,which is not directly related with dimensionality and chemical composition of nanostructure-forming compound or composite and based mainly on the less-common nanostruc-tures,was offered.5This review is focused to the examination of less-common nanostructures (i.e.,published mainly in the range of 1-100reports)in the shape of varieties of vegetation (trees,bushes,grass,ﬂowers,and mushrooms,etc).Such structures,as will be shown below,possess unusual shapes and frequently high surface area,which make them very useful for catalytic and medical applications,among others.Several examples,discussed in this review (for example tree-or broccoli-like structures),possess nanoscale (<100nm)widths and higher (sometimes >500)lengths or heights,which formally exceed nanoscale range.Nanotrees.Nanotrees in General.Tree-like is a relatively frequent nano-and microstructural type,obtained for a series of inorganic compounds,mostly for silicon,silica,and carbon.Generally,they are closely related to nanowires,which are modiﬁed with branches,sometimes they are produced from them,6and can appear together with other nanostructures.In the case of silicon,the structure and electronic properties of its nanotrees using large scale quantum mechanical simulations were investigated.7Electronic structure analysis showed that the formation of nanotrees was accompanied by a narrowing of the HOMO -LUMO gap;this fact could lead to their application in molecular devices due to the predicted enhance-ment in the conducting properties.Si nanotrees were found to be suitable also for thermoelectrical applications.8The effects of trimethylaluminium (TMA)on silicon nanowires grown by CVD were investigated in the 650-850°C growth temperature range using Au as the growth catalyst and SiH 4in H 2carrier gas as the Si precursor.9Depending on substrate temperature and TMA partial pressure,the structure’s morphology evolved from wires to tapered needles,pyramids,or nanotrees (the TMA presence was linked,in particular,to a branched growth leading to Si nanotrees).Hierarchical silicon oxycarbide tree-like nanostructures,which consisted of trunks and abundant branches,were fabricated by electron irradiation of organosilicon polymers.10The nanotree structures had two different morphologies,cluster-assembled and nanowire-assembled,formation of which can be controlled by the irradiation parameters.Superhydrophobic silica nanotrees were obtained by sol -gel method with hybrid silica sol and jelly like resorcinol formaldehyde resin.11The surface roughness of the silica nanotrees ﬁlm is about 20nm,and it is transparent and superhydrophobic with a water contact angle higher than 150°.Single crystal silicon was found to be self-organized into ammonium silicon hexaﬂuoride {(NH 4)2SiF 6}based complex structures under exposure to vapors of an HF-rich HF:HNO 3mixture 12forming structures like pillars or micro and nanotrees*To whom correspondence should be addressed.E-mail:bkhariss@mail.ru.Ind.Eng.Chem.Res.2010,49,11142–111691114210.1021/ie1017139 2010American Chemical SocietyPublished on Web 10/18/2010perpendicular to the wafer plane.It was shown that the reaction of HNO3with(NH4)2SiF6and the H2O vapor plays an important role in the formation of these features.Beautiful porous carbon tree-like nanostructures(Figure1) were obtained13(as well as SiC above)by electron-beam beam-induced deposition method using a TEM in poor vacuum conditions where hydrocarbons(being attracted to the substrates by the local electricﬁelds)were present in the chamber.The authors found that the adequate ion dose to create well-deﬁned sawtooth nanopatterns was between8and10nC/mm2.A high concentration of sp2sites in nanostructures was revealed suggesting that they were made of graphite-like hydrogenated amorphous carbon.Y-junction hollow carbon nanotrees(with close-ended branches but open-ended roots,length∼5µm,an average external trunk diameter of∼500nm,branch diameters ranging from70to100nm,and wall thickness ranging from 24to30nm)were obtained in a90%yield by the reaction of ferrocene with1,2-dichlorobenzene at180°C.14A simple and effective method based on Coulomb effect for separating SWNTs from their bundles was developed.15The separated SWNTs were radially distributed at the end of the original bundle andﬁnally formed a nanotree(Figure2).The formation mechanism was proposed to be Coulomb explosion.As an important application of carbon nanotrees,the fuel cell,equipped with an anode and/or a cathode containing carbon-metal nanotree catalysts consisting of CNT trunks having a plurality of highly oriented carbon nanotube branches,where a metal is attached on inner and/or outer surfaces of the tubes,16provided high power generation efﬁciency by accelerating catalytic functions.Tree-like nanostructures have been also obtained for several transition and noble elemental metals,as well as their oxides and salts.Thus,a metallic tungsten nanotree was formed on an Al2O3substrate by using W(CO)6gas by electron-beam tech-nique.18Platinum metal nanotrees were prepared by assembling Pt nanoparticles onto a DNA network template and reducing metal ions by photoinduced method.19The concentration of DNA network and the reaction time played important roles in forming platinum nanotrees.An environmentally friendly1-step method,with H2O as a benign solvent and vitamin B1as a reducing agent,to synthesize Pd nanobelts,nanoplates,and nanotrees using vitamin B1without using any special capping agents at room temperature was elaborated20and it can be extended to prepare other noble nanomaterials such as Au and Pt.The Pd nanoparticles showed excellent catalytic activity for several C-C bond forming reactions such as Suzuki,Heck,and Sonogashira reactions under microwave irradiation conditions. Palladium nanotrees can also be produced electrochemically; thus,uniform nanostructured Pd(standing Pd nanoplates and nanotrees(Figure3)was electrodeposited directly on Au substrate through a facile template-free approach,which in-volved a cyclic electrochemical deposition/dissolution of Pd.21 Improved and superior electrocatalytic activities toward the oxidation of EtOH were observed from these electrodes coated with nanostructured Pd.A simple route for the synthesis of such highly catalytic Pd nanostructures was proposed.An interesting example of use of TEM grids for nanostructure preparation was reported.Thus,the copper-carbon substrate of a TEM grid reacted with aqueous silver nitrate solution within minutes to yield spectacular tree-like silver dendrites,without using any added capping or reducing reagents.22The authors noted that these results demonstrated a facile,aqueous,room-temperature synthesis of a range of noble metal nano-and meso-structures that have widespread technological potential in the design and development of next-generation fuel cells,catalysts, and antimicrobial coatings.Also,a tree-like nanostructured Ag crystal,with stems,branches,and leaves,was synthesized by preforming Au seeds,soaking,and annealing,based on mono-lithic mesoporous silica.23The obtained Ag nanotrees were of single-crystal nature and statistically symmetric in geometry.A lab-on-a-chip nanotree enzyme reactor,constructed from gold-tipped branched nanorod structures grown on SiN x-covered wafers,was demonstrated for the detection of acetylcholine.24 Signiﬁcantly higher enzymic activity was found for the nanotree reactors than for the nanorod reactors,most likely due to the increased gold surface area and thereby higher enzyme binding capacity.The morphology of Au nanotrees was discussed in ref25.Unusual biaxial textures in Cu nanorodﬁlms,branching into nanotrees,were observed.26Figure 1.Carbon tree-like nanostructures,obtained in porous silicon sample.17Reproduced with permission from ref13.Copyright2009Elsevier Science.Figure2.Typical nanotrees of SWNTs formed at the breaking points of SWNTs bundle.Reproduced with permission from ref15.Copyright2009 American Chemical Society.Figure3.Pd nanotrees(about1µm high and100nm thick).Reproduced with permission from ref21.Copyright2009American Chemical Society.Ind.Eng.Chem.Res.,Vol.49,No.22,201011143Theﬁrst successful attempt(one-pot synthesis approach involving the reaction of a titanium surface with the vapor generated from a hydrochloric acid solution in a Teﬂon lined autoclave)to grow ordered tree-like titania TiO2nanoarrays was reported.27By adjusting the initial HCl concentration,ﬁlms of different rutile structures including nanotrees,dendritic nanobun-dles,and nanorods can be selectively obtained.A possible formation mechanism(Figure4)for the interesting architectures was proposed based on a series of time-dependent experiments. Thinﬁlms of aligned WO3nanotrees were synthesized using a simple hydrothermal reaction again starting from elemental metal(in this case,tungsten).28These WO3nanotrees,composed of trunks and branches,were single crystals oriented in the(001) direction.The thinﬁlms exhibited efﬁcient electrochromism due to their large tunnels in the crystal and nanochannels between the nanotrees.29Magnetite nanoparticles of Fe3O4were obtained by reaction 1and grew into large highly branched nanostructures including nanochains and highly branched nanotrees in the solid state through a postannealing process.30The driving force for the formation of the highly ordered nanostructures includes interac-tion between the nanoparticles and interaction through surface-capping molecules.Gallium oxide(Ga2O3)nanotrees,as well as nanobelts and nanotubes(depending on the temperature zone),were obtained on Si(111)substrates by thermal evaporation of gallium in an ammonia atmosphere.31Photoluminescence spectrum under excitation at396nm showed a green emission,which is probably attributed to vacancies in Ga2O3.ZnO nanotrees(ZnO is one of the most important wide band gap semiconductors with a wide and direct band gap of3.37eV and high binding exciton energy of60meV32)were obtained by different methods;thus,complex ZnO nanotrees were observed,among nanosaws, nanobelts,and nanowires,in the products of spray-pyrolysis-assisted thermal evaporation method of ZnO.33The unique nanotree ZnO arrays(Figure5),obtained in large scales at low cost by direct hydrothermal oxidation of zinc metal to complex ZnO nano/microcrystallite forest,exhibited high photocatalytic activity;this fact may help develop a new suite of branched nanostructures for wurtzite semiconductors in general for a wide variety of important applications.34Microwave absorption properties of ZnO nanotrees were studied.35Among other oxides obtained in nanotree forms,we note Al2O336,37and MgO.38 Metal salts synthesized as tree-like structures belong mainly to sulﬁdes and selenides(CdS and CdSe39),phosphides(GaP,40 GaP/InP41),and arsenides(InAs),i.e.,having important ap-plications as semiconductors.Thus,growth of GaP nanotrees (Figure6)took place when Au aerosol particles were deposited on a substrate with the aid of an electrostaticﬁeld,acting as seeds in the tree growth process.42On each Au seed with a size of40-70nm,a tree trunk of GaP was then grown by a technique called vapor-liquid-solid growth.The nanotree arrays may be applied as light emitters or photovoltaic convert-ers.When the Au-In particles were used to seed a second generation of InAs nanowires,producing nanotrees,the branches exhibited a2-3times higher growth rate and more regular shape than those seeded by pure Au particles.43,44This result was attributed by the authors to the decreased interaction between the seed particle and the trunk nanowires when Au-In particles were used.Nanoscale tree-like FeCl3-S structures were as-sembled45and their unique characteristics,based on the high reactivity of FeCl3moiety and the branched thread morphology of sulfur moiety,endowed them multiple excellent functions, particularly acting as reactive seeds or templates to directly induce organic polymerization reaction and guide the growth of polycyclopentadiene nanoﬁbers.A few organic nanotrees are known.For example,spherical polymer brush particles with“nanotree”-type morphology that consisted of a solid poly(styrene)core(diameter∼100nm)onto which chains of a bottlebrush polymer were densely grafted, were prepared in aqueous dispersion by photo emulsion-polymerization using the macromonomer poly(ethylene glycol) methacrylate(PEGMA)(Figure7).46These particles can be used as nanoreactors for the generation and immobilization of well-deﬁned silver nanoparticles exhibiting an excellent colloidal stability;their catalytic activity was investigated by monitoring the reduction of4-nitrophenol by NaBH4in presence of these silver nanocomposite particles.Figure4.Schematic illustration of the TiO2nanotree growth process:(a) no densely distributed rodlike trunks grow from the titanium and the side surfaces are eroded;(b)obvious branches appear and the trunk tends to be more oriented due to space limitation;(c)both branches and trunks tend to be larger in dimension because of“Ostwald Ripening”;(d)ternary growth of the branches.Reproduced with permission from ref27.Copyright2009 American Chemical Society.Figure5.ZnO nanopine tree.Reproduced with permission from ref34. Copyright2008American Chemical Society.11144Ind.Eng.Chem.Res.,Vol.49,No.22,2010Nanopines.An array of Au nanopine trees was obtained by electron-beam evaporation of a few nanometers of Au on bare alumina templates with pores(50nm wide and500nm long).47 This synthetic approach for producing nanoscaleﬁeld emitters could lead to a versatile and inexpensive technology for synthesizingﬂexible arrays of nanoscale cold cathode emitters. Fe3N nanodendrites were synthesized48directly by reduction-nitriding of R-Fe2O3nanopine dendrites(obtained from K3[Fe(CN)6]by hydrothermal method49)in a mixed stream of H2-NH3.Fe3N basically retained dendritic morphology of the starting material R-Fe2O3(Figure8).In a related investigation,carried out starting from similar precursors but by a different method,nanostructured metal(Fe, Co,Mn,Cr,Mo)oxides were fabricated under microwave irradiation conditions in pure water from inexpensive compounds ([K4Fe(CN)6,K3Co(CN)6,K3Mn(CN)6,K3Mo(CN)8,and K3Cr(CN)6])without using any reducing or capping reagent. The metal oxides self-assembled into pine morphology,among many others(i.e.,octahedra,spheres,triangular rods,and hexagonal snowﬂake-like three-dimensional morphologies).50 After being functionalized and coated with Pd metal(Figure 9),pine-structured nanoiron oxides were studied as a support for various catalytic organic transformations,catalyzing various C-C coupling and hydrogenation reactions with high yields. Wurtzite CdS and CdSe nanostructures with complex mor-phologies including fractal nanotrees(which are a buildup of nanopines)can be produced via a solvothermal approach in a mixed solution made of diethylenetriamine(DETA)and deion-ized water.51The produced CdS nanotrees were found to be made up of many branched nanopines with different growth directions and displayed high hyperbranched nanostructures.In a related report,52the controllable synthesis of hexagonal pine-like Zn-doped CdSe(Cd1-x Zn x Se,x)0.1-0.3)nanotrees from self-prepared mixed-metal precursors was achieved by two facile steps:ﬁrst,mixed-metal precursors(x)0.1-0.3)were prepared directly through precipitation reactions of stoichiometric cad-mium acetate,zinc acetate,and sodium selenite in distilled water under ambient condition;second,pure hexagonal phase pine-like Cd1-x Zn x Se(x)0.1-0.3)nanotrees with different ratios were produced via solvothermal treatment of the precursor compounds in hydrazine hydrate at180°C for12h.The micro-and nanostructures of PbS were synthesized via CVD with PbCl2and elemental sulfur53or H2S54as precursors under argonﬂow with a coﬂow of H2at atmospheric pressure and with temperatures between600and650°C(reactions2-3; reaction4is thermodynamically unfavorable).The effects of various growth parameters,such as hydrogenﬂow,temperature, pressure,reaction time(Figure10),and the growth substrates employed were studied.Nanopalms.Nanopalm-like trees are lesser represented.Silica nanowires(diameters in the range of15-49nm and lengths up to500nm),obtained by electron beam irradiation on porous siliconﬁlms,were exposed to poor vacuum conditions in which carbon aggregation from the surrounding gas was promoted by the local electricﬁelds enhanced at the tip of the silica wires, leading to heterostructures showing a nanopalm-like shape.55,56 ZnO nanopalms were obtained,57among other CuO and ZnO nanostructures,by directly heating a CuZn alloy(brass)on a hot plate in ambient conditions.An individual ZnO nanopalm had a width of about350nm at the root part and separated into several nanowires with tip diameters of about20nm on the upper region.Figure11brieﬂy generalizes the morphological dependence of products on the Zn concentrations in the brasses and growth temperatures.Nanoleaves.Leaf-like nanostructures are known mainly for metal oxides,although carbon,some metals,intermetallicFigure6.(a)TEM shows a tree trunk of gallium phosphide grown on a gold seed,with branches grown from smaller gold particles.(b)SEM shows an ordered array of nanotrees viewed at45°from the normal.Reproduced with permission from ref42.Copyright2004American Institute of Physics. Figure7.Preparation of PS-PEGMA-Ag composite particles:Ag ions are adsorbed by the brush particles in aqueous phase.Subsequent reduction leads to well-deﬁned Ag nanoparticles immobilized in the dense surface layer of the particles.Reproduced with permission from ref46.Copyright2006Elsevier Science.H 2(g)+S(s)f H2S(g)(2)PbCl2(s)+H2S(g)f PbS(s)+2HCl(g)(3)PbCl2(s)+1/2S2(g)f PbS(s)+Cl2(g)(thermodynamically unfavorable)(4)Ind.Eng.Chem.Res.,Vol.49,No.22,201011145compounds,and metal salts have been also reported.Thus,immobilization of Ag nanoleaves on γ-mercaptopropyltrim-ethyoxysilane-modiﬁed chemical vapor deposited diamond ﬁlm was carried out.58Ferromagnetic MnSb ﬁlms containing nanoleaves had a width and thickness of ∼100and 20nm,respectively,were synthesized on Si wafers by physical vapor deposition.592D and 3D carbon nanoﬂowers,prepared on silicon (111)substrates by plasma-enhanced CVD,using CH 4,H 2,and Ar as reactive gases in the presence of Fe catalyst,were formed by various nanoleaves (2D,related closely to the ﬂux ratio of gas and the reaction pressure)or hundreds of nanoﬁbers (3D,depended mainly on the growth temperature).60Cu(OH)2nanoleaves were converted into to 2D CuO nanoleaves (Figure 12)on a large scale by a reconstructive transformation,which consisted of the nucleation of CuO followed by a two-step oriented attachment of the CuO particles to 1D CuO nanoribbons and then to 2D CuO nanoleaves,in which hierarchical-oriented attachment is involved.61In a related publication,62single-crystalline CuO nanoleaves with an average thickness of ∼10nm and with lateral sizes of hundreds of nanometers to several micrometers were synthesized by microwave heating of an aqueous solution containing copper salt and sodium hydroxide.The band gap of CuO nanoleaves was estimated to be ∼2.13eV,which showed signiﬁcant blue-shift compared with that of the bulk.ZnOnanoleaves were synthesized by using a CVD method 63and also by pulsed laser ablation (Nd:YAG 532nm),64in the last case at room temperature and pressure using a Zn target in an aqueous solution of Na dodecyl sulfate.The growth mechanism appeared to involve an increase of the structural complexity from 0-D nanoparticles to 1-D nanorods,and then broadening of these into 2-D “nanoleaf”structures.Alternatively,using a mixture of ZnO and Te powders as the source material,ZnO nanoleaves,con-structed with a nanowire and a nanodisc on one side of the nanowire near the top,with high yield and uniform morphology were fabricated by thermal evaporation.65These ZnO nanoleaves could be used in nanolasers,sensors,and photoelectronic nanode-vices.SnO 2nanoleaves and nanopencils were synthesized 66on single silicon substrates using Au -Ag alloying catalyst assisted carbothermal evaporation.In comparison with SnO 2nanopencils,the new peak at 456nm in the measured PL spectra of SnO 2nanoleaves was observed,implying that more luminescence centers existed in these nanostructures.Freezing ferromagnetic moment model was proposed to explain the exchange bias discovered in R -Fe 2O 3antiferromagnetic nanoleaves 67synthesized by oxygenat-ing pure iron.68The small exchange bias may possibly originate from a different magnetic order on the surface of the nanoleaves or the coexistence of a tiny amount of Fe 3O 4.Single-crystalline TiO 2nanoleaves were functionalized with catechol-group-termi-nated Zn(II)-porphyrin,coordinated with ZnP with trans-2,2′-ethylene 4,4′-bipyridyl,showing the self-assembly of TiO 2nanoleaves based on a supramolecular interaction and formation of a facet-selective,self-assembled,2D stacking structure.69Uniform single-crystal boehmite leaf-like nanosheets with high anisotropy (with a lateral size of 4.5µm ×9.0µm and a thickness of 60-90nm)and ﬂower-like superstructures consisting of single-crystal petals were prepared by a simple hydrothermal method and transformed by calcination into single-crystal gamma-alumina nanostructures while keeping their morphology.70Highly crystallineFigure 8.R -Fe 2O 3(precursor)and Fe 3N (ﬁnal product)pine-like dendrites.Reproduced with permission from ref 48.Copyright 2005Elsevier Science.Figure 9.Synthesis of functionalized ferrites with Pd coating.Reproduced with permission from ref 50.Copyright 2009American Chemical Society.11146Ind.Eng.Chem.Res.,Vol.49,No.22,2010CePO4,CePO4:Tb3+,and CePO4:Dy3+nanoleaves with monoclinic structure and dispersible in solvents such as water and methanol were prepared71by a relatively low temperature(at140°C) synthesis in ethylene glycol medium from Ce2(CO3)3·5H2O,Tb4O7, and Dy2O3as precursors and incorporated into silica sols by the sol-gel method,exhibiting improved luminescence properties compared to silica sols directly doped with lanthanide ions.These systems can be useful for developing biosensors for studying enzyme and protein activities based on luminescent silica sols. Industrially important gallium arsenide and phosphide were also obtained as nanoleaves and characterized.Thus,wurtzitic Ga nitride nano-and microleaves(Figure13)were obtained grown through a CVD method from -Ga2O3as precursor according to the eqs 5-6.72Field effect transistors based on individual GaN nanoleaves were fabricated,and the electric transport results revealed a pronounced n-type gating effect of the GaN nanostructures.GaAs nanowires and nanoleaves were grown by molecular beam epitaxy using Mn as growth catalyst.73Bi2S3nanoleaves,together with other morphologies,including nanorods,dandelion-like nanostruc-tures,nanoﬂowers,and nanocabbages,were synthesized from thesingle-source precursor Bi(SCOPh)3or multiple-source precursorsby using a colloidal solution method or hydrothermal method.74The nanoleaves had an average diameter of89(11nm and lengthof471(52nm and resembled some natural long leaves with awider middle part and two tapering points.Nanoforests.Nanoforests(Figures6b and14),as a con-glomerate of nanotrees above and/or nanobushes and other“nanovegetation”described below,as well as simple nanorodsand nanowires,have been reported mainly for carbon nano-tubes.75A key growth precursor for SWCNTs is acetylene,76although ethylene mixtures with H277(the equipment is shownin Figure15)or C2H4/H2/H2O/Ar78were used,mainly by CVDtechnique.Thus,CNTs nanoforest was prepared79by biasedthermal CVD on Co-containing amorphous C compositeﬁlms.These obtained nanoforest multiwalled CNTs had thin diametersbetween10and20nm.Vertically aligned small diameter(single-and few-walled)carbon nanotube forests were alsogrown by thermal CVD over the temperature range560-800°C and10-5to14mbar partial pressure range,using acetylene as the feedstock and Al2O3-supported Fe nanoparticles as thecatalyst.80Their formation mechanism is described by reactions7-11;an alternative mechanism was also proposed.81Effectsof aluminum buffer layer on synthesis of highly verticallyaligned carbon nanotube(CNT)forests were systematicallystudied.82These CNT forests were synthesized using CVDsystem on iron nanoparticles catalyzed substrate supported bythin aluminum buffer layer with Fe/Al thickness of0.5and40nm,respectively.Yarned CNTﬁbers were spun from grownCNT forests for applications to reinforced materials.Introduction of CO2was found to be a facile way to tune the growth of vertically aligned double-or single-walled carbon nanotube forests on wafers.84In the absence of CO2,a double-walled CNT convexity was obtained;with increasing concentra-tion of CO2,the morphologies of the forests transformedﬁrstFigure10.Progression of PbS nanowire pine trees over reaction time.Reproduced with permission from ref54.Copyright2009American Chemical Society.Figure11.Morphological dependence of products on Zn contents in brass and growth temperatures.The following dominant morphologies are shown: CuO nanowires(CuO NWs),ZnO nanowires(ZnO NWs),ZnO nanoﬂakes (ZnO NFs),and ZnO nanopalms(ZnO NPs).Reproduced with permission from ref57.Copyright2006Wiley Interscience.Figure12.CuO nanoleaves;the upper inset is two single CuO nanoleaves. Reproduced with permission from ref61.Copyright2007American Chemical Society.Ga2O3(s)+4Ga(l)f3Ga2O(g)(5)Ga2O(g)+2NH3(g)f2GaN(s)+H2O(g)+2H2(g)(6)C2H2gasT C2H2ads(7)C2H2adsT2CH ads(8)CHadsT C ads+H ads(9)2HadsT H2gas(10)Cadsf C CNT(11)Ind.Eng.Chem.Res.,Vol.49,No.22,201011147。

T.W. ANDERSON (1971). The Statistical Analysis of Time Series. Series in Probability and Ma

425 BibliographyH.A KAIKE(1974).Markovian representation of stochastic processes and its application to the analysis of autoregressive moving average processes.Annals Institute Statistical Mathematics,vol.26,pp.363-387. B.D.O.A NDERSON and J.B.M OORE(1979).Optimal rmation and System Sciences Series, Prentice Hall,Englewood Cliffs,NJ.T.W.A NDERSON(1971).The Statistical Analysis of Time Series.Series in Probability and Mathematical Statistics,Wiley,New York.R.A NDRE-O BRECHT(1988).A new statistical approach for the automatic segmentation of continuous speech signals.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-36,no1,pp.29-40.R.A NDRE-O BRECHT(1990).Reconnaissance automatique de parole`a partir de segments acoustiques et de mod`e les de Markov cach´e s.Proc.Journ´e es Etude de la Parole,Montr´e al,May1990(in French).R.A NDRE-O BRECHT and H.Y.S U(1988).Three acoustic labellings for phoneme based continuous speech recognition.Proc.Speech’88,Edinburgh,UK,pp.943-950.U.A PPEL and A.VON B RANDT(1983).Adaptive sequential segmentation of piecewise stationary time rmation Sciences,vol.29,no1,pp.27-56.L.A.A ROIAN and H.L EVENE(1950).The effectiveness of quality control procedures.Jal American Statis-tical Association,vol.45,pp.520-529.K.J.A STR¨OM and B.W ITTENMARK(1984).Computer Controlled Systems:Theory and rma-tion and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.M.B AGSHAW and R.A.J OHNSON(1975a).The effect of serial correlation on the performance of CUSUM tests-Part II.Technometrics,vol.17,no1,pp.73-80.M.B AGSHAW and R.A.J OHNSON(1975b).The inﬂuence of reference values and estimated variance on the ARL of CUSUM tests.Jal Royal Statistical Society,vol.37(B),no3,pp.413-420.M.B AGSHAW and R.A.J OHNSON(1977).Sequential procedures for detecting parameter changes in a time-series model.Jal American Statistical Association,vol.72,no359,pp.593-597.R.K.B ANSAL and P.P APANTONI-K AZAKOS(1986).An algorithm for detecting a change in a stochastic process.IEEE rmation Theory,vol.IT-32,no2,pp.227-235.G.A.B ARNARD(1959).Control charts and stochastic processes.Jal Royal Statistical Society,vol.B.21, pp.239-271.A.E.B ASHARINOV andB.S.F LEISHMAN(1962).Methods of the statistical sequential analysis and their radiotechnical applications.Sovetskoe Radio,Moscow(in Russian).M.B ASSEVILLE(1978).D´e viations par rapport au maximum:formules d’arrˆe t et martingales associ´e es. Compte-rendus du S´e minaire de Probabilit´e s,Universit´e de Rennes I.M.B ASSEVILLE(1981).Edge detection using sequential methods for change in level-Part II:Sequential detection of change in mean.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-29,no1,pp.32-50.426B IBLIOGRAPHY M.B ASSEVILLE(1982).A survey of statistical failure detection techniques.In Contribution`a la D´e tectionS´e quentielle de Ruptures de Mod`e les Statistiques,Th`e se d’Etat,Universit´e de Rennes I,France(in English). M.B ASSEVILLE(1986).The two-models approach for the on-line detection of changes in AR processes. In Detection of Abrupt Changes in Signals and Dynamical Systems(M.Basseville,A.Benveniste,eds.). Lecture Notes in Control and Information Sciences,LNCIS77,Springer,New York,pp.169-215.M.B ASSEVILLE(1988).Detecting changes in signals and systems-A survey.Automatica,vol.24,pp.309-326.M.B ASSEVILLE(1989).Distance measures for signal processing and pattern recognition.Signal Process-ing,vol.18,pp.349-369.M.B ASSEVILLE and A.B ENVENISTE(1983a).Design and comparative study of some sequential jump detection algorithms for digital signals.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-31, no3,pp.521-535.M.B ASSEVILLE and A.B ENVENISTE(1983b).Sequential detection of abrupt changes in spectral charac-teristics of digital signals.IEEE rmation Theory,vol.IT-29,no5,pp.709-724.M.B ASSEVILLE and A.B ENVENISTE,eds.(1986).Detection of Abrupt Changes in Signals and Dynamical Systems.Lecture Notes in Control and Information Sciences,LNCIS77,Springer,New York.M.B ASSEVILLE and I.N IKIFOROV(1991).A uniﬁed framework for statistical change detection.Proc.30th IEEE Conference on Decision and Control,Brighton,UK.M.B ASSEVILLE,B.E SPIAU and J.G ASNIER(1981).Edge detection using sequential methods for change in level-Part I:A sequential edge detection algorithm.IEEE Trans.Acoustics,Speech,Signal Processing, vol.ASSP-29,no1,pp.24-31.M.B ASSEVILLE, A.B ENVENISTE and G.M OUSTAKIDES(1986).Detection and diagnosis of abrupt changes in modal characteristics of nonstationary digital signals.IEEE rmation Theory,vol.IT-32,no3,pp.412-417.M.B ASSEVILLE,A.B ENVENISTE,G.M OUSTAKIDES and A.R OUG´E E(1987a).Detection and diagnosis of changes in the eigenstructure of nonstationary multivariable systems.Automatica,vol.23,no3,pp.479-489. M.B ASSEVILLE,A.B ENVENISTE,G.M OUSTAKIDES and A.R OUG´E E(1987b).Optimal sensor location for detecting changes in dynamical behavior.IEEE Trans.Automatic Control,vol.AC-32,no12,pp.1067-1075.M.B ASSEVILLE,A.B ENVENISTE,B.G ACH-D EVAUCHELLE,M.G OURSAT,D.B ONNECASE,P.D OREY, M.P REVOSTO and M.O LAGNON(1993).Damage monitoring in vibration mechanics:issues in diagnos-tics and predictive maintenance.Mechanical Systems and Signal Processing,vol.7,no5,pp.401-423.R.V.B EARD(1971).Failure Accommodation in Linear Systems through Self-reorganization.Ph.D.Thesis, Dept.Aeronautics and Astronautics,MIT,Cambridge,MA.A.B ENVENISTE and J.J.F UCHS(1985).Single sample modal identiﬁcation of a nonstationary stochastic process.IEEE Trans.Automatic Control,vol.AC-30,no1,pp.66-74.A.B ENVENISTE,M.B ASSEVILLE and G.M OUSTAKIDES(1987).The asymptotic local approach to change detection and model validation.IEEE Trans.Automatic Control,vol.AC-32,no7,pp.583-592.A.B ENVENISTE,M.M ETIVIER and P.P RIOURET(1990).Adaptive Algorithms and Stochastic Approxima-tions.Series on Applications of Mathematics,(A.V.Balakrishnan,I.Karatzas,M.Yor,eds.).Springer,New York.A.B ENVENISTE,M.B ASSEVILLE,L.E L G HAOUI,R.N IKOUKHAH and A.S.W ILLSKY(1992).An optimum robust approach to statistical failure detection and identiﬁcation.IFAC World Conference,Sydney, July1993.B IBLIOGRAPHY427 R.H.B ERK(1973).Some asymptotic aspects of sequential analysis.Annals Statistics,vol.1,no6,pp.1126-1138.R.H.B ERK(1975).Locally most powerful sequential test.Annals Statistics,vol.3,no2,pp.373-381.P.B ILLINGSLEY(1968).Convergence of Probability Measures.Wiley,New York.A.F.B ISSELL(1969).Cusum techniques for quality control.Applied Statistics,vol.18,pp.1-30.M.E.B IVAIKOV(1991).Control of the sample size for recursive estimation of parameters subject to abrupt changes.Automation and Remote Control,no9,pp.96-103.R.E.B LAHUT(1987).Principles and Practice of Information Theory.Addison-Wesley,Reading,MA.I.F.B LAKE and W.C.L INDSEY(1973).Level-crossing problems for random processes.IEEE r-mation Theory,vol.IT-19,no3,pp.295-315.G.B ODENSTEIN and H.M.P RAETORIUS(1977).Feature extraction from the encephalogram by adaptive segmentation.Proc.IEEE,vol.65,pp.642-652.T.B OHLIN(1977).Analysis of EEG signals with changing spectra using a short word Kalman estimator. Mathematical Biosciences,vol.35,pp.221-259.W.B¨OHM and P.H ACKL(1990).Improved bounds for the average run length of control charts based on ﬁnite weighted sums.Annals Statistics,vol.18,no4,pp.1895-1899.T.B OJDECKI and J.H OSZA(1984).On a generalized disorder problem.Stochastic Processes and their Applications,vol.18,pp.349-359.L.I.B ORODKIN and V.V.M OTTL’(1976).Algorithm forﬁnding the jump times of random process equation parameters.Automation and Remote Control,vol.37,no6,Part1,pp.23-32.A.A.B OROVKOV(1984).Theory of Mathematical Statistics-Estimation and Hypotheses Testing,Naouka, Moscow(in Russian).Translated in French under the title Statistique Math´e matique-Estimation et Tests d’Hypoth`e ses,Mir,Paris,1987.G.E.P.B OX and G.M.J ENKINS(1970).Time Series Analysis,Forecasting and Control.Series in Time Series Analysis,Holden-Day,San Francisco.A.VON B RANDT(1983).Detecting and estimating parameters jumps using ladder algorithms and likelihood ratio test.Proc.ICASSP,Boston,MA,pp.1017-1020.A.VON B RANDT(1984).Modellierung von Signalen mit Sprunghaft Ver¨a nderlichem Leistungsspektrum durch Adaptive Segmentierung.Doctor-Engineer Dissertation,M¨u nchen,RFA(in German).S.B RAUN,ed.(1986).Mechanical Signature Analysis-Theory and Applications.Academic Press,London. L.B REIMAN(1968).Probability.Series in Statistics,Addison-Wesley,Reading,MA.G.S.B RITOV and L.A.M IRONOVSKI(1972).Diagnostics of linear systems of automatic regulation.Tekh. Kibernetics,vol.1,pp.76-83.B.E.B RODSKIY and B.S.D ARKHOVSKIY(1992).Nonparametric Methods in Change-point Problems. Kluwer Academic,Boston.L.D.B ROEMELING(1982).Jal Econometrics,vol.19,Special issue on structural change in Econometrics. L.D.B ROEMELING and H.T SURUMI(1987).Econometrics and Structural Change.Dekker,New York. D.B ROOK and D.A.E VANS(1972).An approach to the probability distribution of Cusum run length. Biometrika,vol.59,pp.539-550.J.B RUNET,D.J AUME,M.L ABARR`E RE,A.R AULT and M.V ERG´E(1990).D´e tection et Diagnostic de Pannes.Trait´e des Nouvelles Technologies,S´e rie Diagnostic et Maintenance,Herm`e s,Paris(in French).428B IBLIOGRAPHY S.P.B RUZZONE and M.K AVEH(1984).Information tradeoffs in using the sample autocorrelation function in ARMA parameter estimation.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-32,no4, pp.701-715.A.K.C AGLAYAN(1980).Necessary and sufﬁcient conditions for detectability of jumps in linear systems. IEEE Trans.Automatic Control,vol.AC-25,no4,pp.833-834.A.K.C AGLAYAN and R.E.L ANCRAFT(1983).Reinitialization issues in fault tolerant systems.Proc.Amer-ican Control Conf.,pp.952-955.A.K.C AGLAYAN,S.M.A LLEN and K.W EHMULLER(1988).Evaluation of a second generation reconﬁgu-ration strategy for aircraftﬂight control systems subjected to actuator failure/surface damage.Proc.National Aerospace and Electronic Conference,Dayton,OH.P.E.C AINES(1988).Linear Stochastic Systems.Series in Probability and Mathematical Statistics,Wiley, New York.M.J.C HEN and J.P.N ORTON(1987).Estimation techniques for tracking rapid parameter changes.Intern. Jal Control,vol.45,no4,pp.1387-1398.W.K.C HIU(1974).The economic design of cusum charts for controlling normal mean.Applied Statistics, vol.23,no3,pp.420-433.E.Y.C HOW(1980).A Failure Detection System Design Methodology.Ph.D.Thesis,M.I.T.,L.I.D.S.,Cam-bridge,MA.E.Y.C HOW and A.S.W ILLSKY(1984).Analytical redundancy and the design of robust failure detection systems.IEEE Trans.Automatic Control,vol.AC-29,no3,pp.689-691.Y.S.C HOW,H.R OBBINS and D.S IEGMUND(1971).Great Expectations:The Theory of Optimal Stop-ping.Houghton-Mifﬂin,Boston.R.N.C LARK,D.C.F OSTH and V.M.W ALTON(1975).Detection of instrument malfunctions in control systems.IEEE Trans.Aerospace Electronic Systems,vol.AES-11,pp.465-473.A.C OHEN(1987).Biomedical Signal Processing-vol.1:Time and Frequency Domain Analysis;vol.2: Compression and Automatic Recognition.CRC Press,Boca Raton,FL.J.C ORGE and F.P UECH(1986).Analyse du rythme cardiaque foetal par des m´e thodes de d´e tection de ruptures.Proc.7th INRIA Int.Conf.Analysis and optimization of Systems.Antibes,FR(in French).D.R.C OX and D.V.H INKLEY(1986).Theoretical Statistics.Chapman and Hall,New York.D.R.C OX and H.D.M ILLER(1965).The Theory of Stochastic Processes.Wiley,New York.S.V.C ROWDER(1987).A simple method for studying run-length distributions of exponentially weighted moving average charts.Technometrics,vol.29,no4,pp.401-407.H.C S¨ORG¨O and L.H ORV´ATH(1988).Nonparametric methods for change point problems.In Handbook of Statistics(P.R.Krishnaiah,C.R.Rao,eds.),vol.7,Elsevier,New York,pp.403-425.R.B.D AVIES(1973).Asymptotic inference in stationary gaussian time series.Advances Applied Probability, vol.5,no3,pp.469-497.J.C.D ECKERT,M.N.D ESAI,J.J.D EYST and A.S.W ILLSKY(1977).F-8DFBW sensor failure identiﬁcation using analytical redundancy.IEEE Trans.Automatic Control,vol.AC-22,no5,pp.795-803.M.H.D E G ROOT(1970).Optimal Statistical Decisions.Series in Probability and Statistics,McGraw-Hill, New York.J.D ESHAYES and D.P ICARD(1979).Tests de ruptures dans un mod`e pte-Rendus de l’Acad´e mie des Sciences,vol.288,Ser.A,pp.563-566(in French).B IBLIOGRAPHY429 J.D ESHAYES and D.P ICARD(1983).Ruptures de Mod`e les en Statistique.Th`e ses d’Etat,Universit´e deParis-Sud,Orsay,France(in French).J.D ESHAYES and D.P ICARD(1986).Off-line statistical analysis of change-point models using non para-metric and likelihood methods.In Detection of Abrupt Changes in Signals and Dynamical Systems(M. Basseville,A.Benveniste,eds.).Lecture Notes in Control and Information Sciences,LNCIS77,Springer, New York,pp.103-168.B.D EVAUCHELLE-G ACH(1991).Diagnostic M´e canique des Fatigues sur les Structures Soumises`a des Vibrations en Ambiance de Travail.Th`e se de l’Universit´e Paris IX Dauphine(in French).B.D EVAUCHELLE-G ACH,M.B ASSEVILLE and A.B ENVENISTE(1991).Diagnosing mechanical changes in vibrating systems.Proc.SAFEPROCESS’91,Baden-Baden,FRG,pp.85-89.R.D I F RANCESCO(1990).Real-time speech segmentation using pitch and convexity jump models:applica-tion to variable rate speech coding.IEEE Trans.Acoustics,Speech,Signal Processing,vol.ASSP-38,no5, pp.741-748.X.D ING and P.M.F RANK(1990).Fault detection via factorization approach.Systems and Control Letters, vol.14,pp.431-436.J.L.D OOB(1953).Stochastic Processes.Wiley,New York.V.D RAGALIN(1988).Asymptotic solutions in detecting a change in distribution under an unknown param-eter.Statistical Problems of Control,Issue83,Vilnius,pp.45-52.B.D UBUISSON(1990).Diagnostic et Reconnaissance des Formes.Trait´e des Nouvelles Technologies,S´e rie Diagnostic et Maintenance,Herm`e s,Paris(in French).A.J.D UNCAN(1986).Quality Control and Industrial Statistics,5th edition.Richard D.Irwin,Inc.,Home-wood,IL.J.D URBIN(1971).Boundary-crossing probabilities for the Brownian motion and Poisson processes and techniques for computing the power of the Kolmogorov-Smirnov test.Jal Applied Probability,vol.8,pp.431-453.J.D URBIN(1985).Theﬁrst passage density of the crossing of a continuous Gaussian process to a general boundary.Jal Applied Probability,vol.22,no1,pp.99-122.A.E MAMI-N AEINI,M.M.A KHTER and S.M.R OCK(1988).Effect of model uncertainty on failure detec-tion:the threshold selector.IEEE Trans.Automatic Control,vol.AC-33,no12,pp.1106-1115.J.D.E SARY,F.P ROSCHAN and D.W.W ALKUP(1967).Association of random variables with applications. Annals Mathematical Statistics,vol.38,pp.1466-1474.W.D.E WAN and K.W.K EMP(1960).Sampling inspection of continuous processes with no autocorrelation between successive results.Biometrika,vol.47,pp.263-280.G.F AVIER and A.S MOLDERS(1984).Adaptive smoother-predictors for tracking maneuvering targets.Proc. 23rd Conf.Decision and Control,Las Vegas,NV,pp.831-836.W.F ELLER(1966).An Introduction to Probability Theory and Its Applications,vol.2.Series in Probability and Mathematical Statistics,Wiley,New York.R.A.F ISHER(1925).Theory of statistical estimation.Proc.Cambridge Philosophical Society,vol.22, pp.700-725.M.F ISHMAN(1988).Optimization of the algorithm for the detection of a disorder,based on the statistic of exponential smoothing.In Statistical Problems of Control,Issue83,Vilnius,pp.146-151.R.F LETCHER(1980).Practical Methods of Optimization,2volumes.Wiley,New York.P.M.F RANK(1990).Fault diagnosis in dynamic systems using analytical and knowledge based redundancy -A survey and new results.Automatica,vol.26,pp.459-474.430B IBLIOGRAPHY P.M.F RANK(1991).Enhancement of robustness in observer-based fault detection.Proc.SAFEPRO-CESS’91,Baden-Baden,FRG,pp.275-287.P.M.F RANK and J.W¨UNNENBERG(1989).Robust fault diagnosis using unknown input observer schemes. In Fault Diagnosis in Dynamic Systems-Theory and Application(R.Patton,P.Frank,R.Clark,eds.). International Series in Systems and Control Engineering,Prentice Hall International,London,UK,pp.47-98.K.F UKUNAGA(1990).Introduction to Statistical Pattern Recognition,2d ed.Academic Press,New York. S.I.G ASS(1958).Linear Programming:Methods and Applications.McGraw Hill,New York.W.G E and C.Z.F ANG(1989).Extended robust observation approach for failure isolation.Int.Jal Control, vol.49,no5,pp.1537-1553.W.G ERSCH(1986).Two applications of parametric time series modeling methods.In Mechanical Signature Analysis-Theory and Applications(S.Braun,ed.),chap.10.Academic Press,London.J.J.G ERTLER(1988).Survey of model-based failure detection and isolation in complex plants.IEEE Control Systems Magazine,vol.8,no6,pp.3-11.J.J.G ERTLER(1991).Analytical redundancy methods in fault detection and isolation.Proc.SAFEPRO-CESS’91,Baden-Baden,FRG,pp.9-22.B.K.G HOSH(1970).Sequential Tests of Statistical Hypotheses.Addison-Wesley,Cambridge,MA.I.N.G IBRA(1975).Recent developments in control charts techniques.Jal Quality Technology,vol.7, pp.183-192.J.P.G ILMORE and R.A.M C K ERN(1972).A redundant strapdown inertial reference unit(SIRU).Jal Space-craft,vol.9,pp.39-47.M.A.G IRSHICK and H.R UBIN(1952).A Bayes approach to a quality control model.Annals Mathematical Statistics,vol.23,pp.114-125.A.L.G OEL and S.M.W U(1971).Determination of the ARL and a contour nomogram for CUSUM charts to control normal mean.Technometrics,vol.13,no2,pp.221-230.P.L.G OLDSMITH and H.W HITFIELD(1961).Average run lengths in cumulative chart quality control schemes.Technometrics,vol.3,pp.11-20.G.C.G OODWIN and K.S.S IN(1984).Adaptive Filtering,Prediction and rmation and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.R.M.G RAY and L.D.D AVISSON(1986).Random Processes:a Mathematical Approach for Engineers. Information and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.C.G UEGUEN and L.L.S CHARF(1980).Exact maximum likelihood identiﬁcation for ARMA models:a signal processing perspective.Proc.1st EUSIPCO,Lausanne.D.E.G USTAFSON, A.S.W ILLSKY,J.Y.W ANG,M.C.L ANCASTER and J.H.T RIEBWASSER(1978). ECG/VCG rhythm diagnosis using statistical signal analysis.Part I:Identiﬁcation of persistent rhythms. Part II:Identiﬁcation of transient rhythms.IEEE Trans.Biomedical Engineering,vol.BME-25,pp.344-353 and353-361.F.G USTAFSSON(1991).Optimal segmentation of linear regression parameters.Proc.IFAC/IFORS Symp. Identiﬁcation and System Parameter Estimation,Budapest,pp.225-229.T.H¨AGGLUND(1983).New Estimation Techniques for Adaptive Control.Ph.D.Thesis,Lund Institute of Technology,Lund,Sweden.T.H¨AGGLUND(1984).Adaptive control of systems subject to large parameter changes.Proc.IFAC9th World Congress,Budapest.B IBLIOGRAPHY431 P.H ALL and C.C.H EYDE(1980).Martingale Limit Theory and its Application.Probability and Mathemat-ical Statistics,a Series of Monographs and Textbooks,Academic Press,New York.W.J.H ALL,R.A.W IJSMAN and J.K.G HOSH(1965).The relationship between sufﬁciency and invariance with applications in sequential analysis.Ann.Math.Statist.,vol.36,pp.576-614.E.J.H ANNAN and M.D EISTLER(1988).The Statistical Theory of Linear Systems.Series in Probability and Mathematical Statistics,Wiley,New York.J.D.H EALY(1987).A note on multivariate CuSum procedures.Technometrics,vol.29,pp.402-412.D.M.H IMMELBLAU(1970).Process Analysis by Statistical Methods.Wiley,New York.D.M.H IMMELBLAU(1978).Fault Detection and Diagnosis in Chemical and Petrochemical Processes. Chemical Engineering Monographs,vol.8,Elsevier,Amsterdam.W.G.S.H INES(1976a).A simple monitor of a system with sudden parameter changes.IEEE r-mation Theory,vol.IT-22,no2,pp.210-216.W.G.S.H INES(1976b).Improving a simple monitor of a system with sudden parameter changes.IEEE rmation Theory,vol.IT-22,no4,pp.496-499.D.V.H INKLEY(1969).Inference about the intersection in two-phase regression.Biometrika,vol.56,no3, pp.495-504.D.V.H INKLEY(1970).Inference about the change point in a sequence of random variables.Biometrika, vol.57,no1,pp.1-17.D.V.H INKLEY(1971).Inference about the change point from cumulative sum-tests.Biometrika,vol.58, no3,pp.509-523.D.V.H INKLEY(1971).Inference in two-phase regression.Jal American Statistical Association,vol.66, no336,pp.736-743.J.R.H UDDLE(1983).Inertial navigation system error-model considerations in Kalmanﬁltering applica-tions.In Control and Dynamic Systems(C.T.Leondes,ed.),Academic Press,New York,pp.293-339.J.S.H UNTER(1986).The exponentially weighted moving average.Jal Quality Technology,vol.18,pp.203-210.I.A.I BRAGIMOV and R.Z.K HASMINSKII(1981).Statistical Estimation-Asymptotic Theory.Applications of Mathematics Series,vol.16.Springer,New York.R.I SERMANN(1984).Process fault detection based on modeling and estimation methods-A survey.Auto-matica,vol.20,pp.387-404.N.I SHII,A.I WATA and N.S UZUMURA(1979).Segmentation of nonstationary time series.Int.Jal Systems Sciences,vol.10,pp.883-894.J.E.J ACKSON and R.A.B RADLEY(1961).Sequential and tests.Annals Mathematical Statistics, vol.32,pp.1063-1077.B.J AMES,K.L.J AMES and D.S IEGMUND(1988).Conditional boundary crossing probabilities with appli-cations to change-point problems.Annals Probability,vol.16,pp.825-839.M.K.J EERAGE(1990).Reliability analysis of fault-tolerant IMU architectures with redundant inertial sen-sors.IEEE Trans.Aerospace and Electronic Systems,vol.AES-5,no.7,pp.23-27.N.L.J OHNSON(1961).A simple theoretical approach to cumulative sum control charts.Jal American Sta-tistical Association,vol.56,pp.835-840.N.L.J OHNSON and F.C.L EONE(1962).Cumulative sum control charts:mathematical principles applied to their construction and use.Parts I,II,III.Industrial Quality Control,vol.18,pp.15-21;vol.19,pp.29-36; vol.20,pp.22-28.432B IBLIOGRAPHY R.A.J OHNSON and M.B AGSHAW(1974).The effect of serial correlation on the performance of CUSUM tests-Part I.Technometrics,vol.16,no.1,pp.103-112.H.L.J ONES(1973).Failure Detection in Linear Systems.Ph.D.Thesis,Dept.Aeronautics and Astronautics, MIT,Cambridge,MA.R.H.J ONES,D.H.C ROWELL and L.E.K APUNIAI(1970).Change detection model for serially correlated multivariate data.Biometrics,vol.26,no2,pp.269-280.M.J URGUTIS(1984).Comparison of the statistical properties of the estimates of the change times in an autoregressive process.In Statistical Problems of Control,Issue65,Vilnius,pp.234-243(in Russian).T.K AILATH(1980).Linear rmation and System Sciences Series,Prentice Hall,Englewood Cliffs,NJ.L.V.K ANTOROVICH and V.I.K RILOV(1958).Approximate Methods of Higher Analysis.Interscience,New York.S.K ARLIN and H.M.T AYLOR(1975).A First Course in Stochastic Processes,2d ed.Academic Press,New York.S.K ARLIN and H.M.T AYLOR(1981).A Second Course in Stochastic Processes.Academic Press,New York.D.K AZAKOS and P.P APANTONI-K AZAKOS(1980).Spectral distance measures between gaussian pro-cesses.IEEE Trans.Automatic Control,vol.AC-25,no5,pp.950-959.K.W.K EMP(1958).Formula for calculating the operating characteristic and average sample number of some sequential tests.Jal Royal Statistical Society,vol.B-20,no2,pp.379-386.K.W.K EMP(1961).The average run length of the cumulative sum chart when a V-mask is used.Jal Royal Statistical Society,vol.B-23,pp.149-153.K.W.K EMP(1967a).Formal expressions which can be used for the determination of operating character-istics and average sample number of a simple sequential test.Jal Royal Statistical Society,vol.B-29,no2, pp.248-262.K.W.K EMP(1967b).A simple procedure for determining upper and lower limits for the average sample run length of a cumulative sum scheme.Jal Royal Statistical Society,vol.B-29,no2,pp.263-265.D.P.K ENNEDY(1976).Some martingales related to cumulative sum tests and single server queues.Stochas-tic Processes and Appl.,vol.4,pp.261-269.T.H.K ERR(1980).Statistical analysis of two-ellipsoid overlap test for real time failure detection.IEEE Trans.Automatic Control,vol.AC-25,no4,pp.762-772.T.H.K ERR(1982).False alarm and correct detection probabilities over a time interval for restricted classes of failure detection algorithms.IEEE rmation Theory,vol.IT-24,pp.619-631.T.H.K ERR(1987).Decentralizedﬁltering and redundancy management for multisensor navigation.IEEE Trans.Aerospace and Electronic systems,vol.AES-23,pp.83-119.Minor corrections on p.412and p.599 (May and July issues,respectively).R.A.K HAN(1978).Wald’s approximations to the average run length in cusum procedures.Jal Statistical Planning and Inference,vol.2,no1,pp.63-77.R.A.K HAN(1979).Someﬁrst passage problems related to cusum procedures.Stochastic Processes and Applications,vol.9,no2,pp.207-215.R.A.K HAN(1981).A note on Page’s two-sided cumulative sum procedures.Biometrika,vol.68,no3, pp.717-719.B IBLIOGRAPHY433 V.K IREICHIKOV,V.M ANGUSHEV and I.N IKIFOROV(1990).Investigation and application of CUSUM algorithms to monitoring of sensors.In Statistical Problems of Control,Issue89,Vilnius,pp.124-130(in Russian).G.K ITAGAWA and W.G ERSCH(1985).A smoothness prior time-varying AR coefﬁcient modeling of non-stationary covariance time series.IEEE Trans.Automatic Control,vol.AC-30,no1,pp.48-56.N.K LIGIENE(1980).Probabilities of deviations of the change point estimate in statistical models.In Sta-tistical Problems of Control,Issue83,Vilnius,pp.80-86(in Russian).N.K LIGIENE and L.T ELKSNYS(1983).Methods of detecting instants of change of random process prop-erties.Automation and Remote Control,vol.44,no10,Part II,pp.1241-1283.J.K ORN,S.W.G ULLY and A.S.W ILLSKY(1982).Application of the generalized likelihood ratio algorithm to maneuver detection and estimation.Proc.American Control Conf.,Arlington,V A,pp.792-798.P.R.K RISHNAIAH and B.Q.M IAO(1988).Review about estimation of change points.In Handbook of Statistics(P.R.Krishnaiah,C.R.Rao,eds.),vol.7,Elsevier,New York,pp.375-402.P.K UDVA,N.V ISWANADHAM and A.R AMAKRISHNAN(1980).Observers for linear systems with unknown inputs.IEEE Trans.Automatic Control,vol.AC-25,no1,pp.113-115.S.K ULLBACK(1959).Information Theory and Statistics.Wiley,New York(also Dover,New York,1968). K.K UMAMARU,S.S AGARA and T.S¨ODERSTR¨OM(1989).Some statistical methods for fault diagnosis for dynamical systems.In Fault Diagnosis in Dynamic Systems-Theory and Application(R.Patton,P.Frank,R. Clark,eds.).International Series in Systems and Control Engineering,Prentice Hall International,London, UK,pp.439-476.A.K USHNIR,I.N IKIFOROV and I.S AVIN(1983).Statistical adaptive algorithms for automatic detection of seismic signals-Part I:One-dimensional case.In Earthquake Prediction and the Study of the Earth Structure,Naouka,Moscow(Computational Seismology,vol.15),pp.154-159(in Russian).L.L ADELLI(1990).Diffusion approximation for a pseudo-likelihood test process with application to de-tection of change in stochastic system.Stochastics and Stochastics Reports,vol.32,pp.1-25.T.L.L A¨I(1974).Control charts based on weighted sums.Annals Statistics,vol.2,no1,pp.134-147.T.L.L A¨I(1981).Asymptotic optimality of invariant sequential probability ratio tests.Annals Statistics, vol.9,no2,pp.318-333.D.G.L AINIOTIS(1971).Joint detection,estimation,and system identiﬁrmation and Control, vol.19,pp.75-92.M.R.L EADBETTER,G.L INDGREN and H.R OOTZEN(1983).Extremes and Related Properties of Random Sequences and Processes.Series in Statistics,Springer,New York.L.L E C AM(1960).Locally asymptotically normal families of distributions.Univ.California Publications in Statistics,vol.3,pp.37-98.L.L E C AM(1986).Asymptotic Methods in Statistical Decision Theory.Series in Statistics,Springer,New York.E.L.L EHMANN(1986).Testing Statistical Hypotheses,2d ed.Wiley,New York.J.P.L EHOCZKY(1977).Formulas for stopped diffusion processes with stopping times based on the maxi-mum.Annals Probability,vol.5,no4,pp.601-607.H.R.L ERCHE(1980).Boundary Crossing of Brownian Motion.Lecture Notes in Statistics,vol.40,Springer, New York.L.L JUNG(1987).System Identiﬁcation-Theory for the rmation and System Sciences Series, Prentice Hall,Englewood Cliffs,NJ.。

metallurgicallaboratory

METALLURGICAL INSPECTIONMetallurgical inspection of the component(s) consists of one or more of the following, as required by the TDP, STDP, engineering drawings, relevant engineering specifications, and AMCOM direction. The results of the metallurgical i n spection are shown in Table 7.x, where x denotes a particular component (e.g. the table for the first component is denoted Table 7.1 and the table for the second component is denoted Table 7.2). The data sheets and photographs are included in the applicable component enclosure.Shot Peening – Shot peening evaluation is performed by qualitatively evaluating shot peening characteristics such as dimple size, degree of coverage, dimple overlap, overspray, and masking. As required, either the entire component surface or selected regions are inspected. A low-power stereomicroscope (up to 50X magnification) is used for the evaluation and the typical appearance is recorded via a high-resolution digital imaging system.Electrical Conductivity– Electrical conductivity is inductively measured using the eddy current technique to verify the heat treatment (temper) of aluminum and magnesium alloys.Nital Etch – Nital etch inspection is typically performed by chemically etching the component bare surface with a solution of nitric acid (2 to 4%) in alcohol (ethanol or methanol). The etched surface is examined visually to detect detrimental microstructural modifications (e.g. untempered martensite) that result from overheating during improper machining (e.g. surface grinding). Other solutions may be used depending on requirements. The inspection is also known as “Temper Etch”. Bulk (Average) Hardness – Hardness is measured by using standard or superficial hardness testers at the locations specified by the TDP. Where measurement locations are not specified by the TDP, the DTB Engineering Department will identify appropriate locations for the tests. For case hardened components, the hardness is measured separately for the case and the core. For very small components, the hardness is measured on a metallographic section by using a microhardness tester. Bulk (Average) Composition – Composition is measured by using an arc emission spectrometer or Inductively Coupled Plasma – Atomic Emission Spectrometer (ICP-AES). Where required, Energy Dispersive Spectroscopy (EDS) is used to supplement the results.Microstructure – Microstructure is evaluated by using standard metallographic techniques including sectioning, mounting, polishing, chemical etching, and inverted optical microscopy in bright field or polarized light as necessary. When required, the analysis is supplemented by scanning electron microscopy. Additional metallurgical inspections are performed only when required by the TDP. It should be noted that the particular technique(s) used for the microstructural analysis of a component may vary with respect to the type and sequence used.Heat Treatment – Heat treatment is checked by coupled microstructural evaluation and hardness measurements, and supplemented by electrical conductivity measurements when necessary. Inclusions – Inclusions are checked by using appropriate metallographic sections to analyze inclusion type and amount (volume fraction or length). When required, composition of the inclusions is checked by EDS.Plating/Coating – Composition and thickness are checked using an appropriate metallographic section that includes a representative plated or coated surface. Plating or coating thickness is measured on the metallographically prepared section using a calibrated digital measurement system. Composition is analyzed by a combination of scanning electron microscopy and energy dispersive spectroscopy.Grain Size – Grain structure is revealed by appropriate chemical etchants and measured by comparison using a microscope eyepiece equipped with the standard ASTM grain size grid. When necessary (e.g. for very fine or coarse grain sizes outside the range of the grid), standard stereological measurements are performed.Grain Flow – Grain flow is determined by using a section(s) of appropriate size and orientation. The section is prepared using standard metallographic techniques, and chemically etched to reveal the grain structure and deformation patterns. The grain flow is analyzed with respect to the component geometry and photographically recorded using either an inverted optical metallograph or macrolevel high-resolution digital imaging.Decarburization– Decarburization is checked by using appropriate metallographic sections that include a representative portion of the component su rface. Chemical etching is used to reveal the presence of decarburization. If decarburization is detected, the depth is measured using a calibrated digital image analysis system and compared with the requirements. In addition, if the decarburized layer is of sufficient depth, microhardness measurements are used as a confirmatory test.Retained Austenite – Retained austenite is revealed by special chemical etching of standard metallographic sections. Microstructural estimation of retained austenite content i s performed by standard stereological volume fraction counts. If the retained austenite content is found to exceed 3% by volume, X-ray diffraction is used as a confirmatory test and the results provided for comparison. Case Hardening – Case hardening is ch ecked by using an appropriate metallographic section that includes a representative portion of the component surface. The “case” is revealed by an appropriate chemical etchant and measured by a microhardness scan from the surface to the interior. Typically, the case depth is reported as the distance from the surface at which the hardness is HRC 50.Table 7.1 – Results of Metallurgical Analysis of Rod End Assembly (Rod End),Part No. 114CS123-1 InspectionSpecification Requirement Actual Reading Findings Photo No. CC (4) Shot Peening Drawing No. 114CS123, Notes 3, 4, 13 and 18, MS21.01* 200% Coverage on Rod End bore Satisfactory, Rod Endbore shot peened withuniform intensity andcoverage > 100%. Cdpeaks were detected byEDS inside the bore,with a plating thicknessof .00030.N/A* 2.1 4 Bulk Hardness Drawing No. 114CS123,Notes 11 and19HRC 39 – 42 HRC 42.0, 42.0, 42.0Meets requirements. No anomalies were noted. --- 4 Average Composition AMS-S-5000 4340 Steel 4340 Steel. See analysisreport.Meets requirements. No anomalies were noted. --- Grain Flow Drawing No. 114CS123, Note 13 and ADCN No. 9 Rolled Threads Satisfactory for RolledThreads. Seerepresentativephotograph.Meets requirements. No anomalies were noted. 2.2 Microstructure AMS-S-5000 4340 Steel Nominal for 4340 Steel,microstructure showsuniformly temperedmartensite.Nominally meets requirements. No anomalies were noted. 2.3 Heat Treatment Drawing No. 114CS123, Note 11 HRC 39 – 42 Satisfactory based onhardness andmicrostructure.Meets requirements. No anomalies were noted. --- Grain Size AMS-S-5000 ASTM No. 5 or finer withoccasionalgrains as largeas ASTM No. 3Finer than ASTM No. 5.Meets requirements. No anomalies were noted. --- Decarburization AMS-S-5000 Section 3.6 Satisfactory. Nodecarburization wasdetected.Meets requirements. No anomalies were noted. --- Cadmium Plating (Composition) A qualitative EDS wasperformed. Cd wasdetected both on theRod End surface andbore.Nominally meets requirements. No anomalies were noted. --- ** Plating/Coating AMS-QQ-P-416 Check and Report, typically .0003 - .0008 (Thickness) The Cadmium Platingmeasured .00048 on theRod End surface and.00030 on the Rod Endbore. No anomalies were noted. 2.4 2.5*Specification was not available.**Baking after cadmium plating is a <<CC>>; it cannot be verified.Table 7.2 – Results of Metallurgical Analysis of Rod End Assembly (Bearing),Part No. 114CS123-1Inspection Specification Requirement Actual Reading Findings PhotoNo.CC(4)Bulk Hardness NoneListed Check andreportBall: HRC 58.0, 58.0,58.0Race: HR15N 79.0,80.0, 80.0 ≅ HRC 38.5N/A ---Average Composition NoneListedCheck andreportBall: 17-4 PH CRESper AMS-QQ-S-763Race: 17-4 PH CRESper AMS 5643N/A ---。

微纳米点阵力学超材料的设计和性能

第43卷第1期力学与实践2021年2月微纳米点阵力学超材料的设计和性能张璇李晓雁t莱布尼茨新材料研宂所，德国萨尔布吕肯66123)t(清华大学工程力学系先进力学与材料中心，北京100084)摘要在过去几年中，增材制造技术的不断涌现促进了点阵力学超材料研究的迅猛发展。

本文主要介绍三维微纳米点阵力学超材料的性能和设计，强调点阵单胞基本单元的演化发展：从桁架到平板再到曲面，如何降低结构内部的应力集中，一步步实现理论预测的各向同性的H a sh in-S h tr ik m a n极限刚度。

并对这一类力学超材料在生物医药、能源环境等领域的应用进行概述。

关键词点阵力学超材料，超高强度，各向同性，H ashh卜S h tr ik m a n极限刚度中图分类号：T B12 文献标识码：A d o i:10.6052/1000-0879-20-100D E S IG N A N D M E C H A N IC A L P R O P E R T Y OF M IC R O-/N A N O-S C A L EA R C H IT E C T E D M ET A M A T E R IA L SZHANG Xuan*，1)LI Xiaoyant* (Leibniz In stitu te for New M aterials, Saarbriicken 66123, G erm any )"^(Department of Engineering M echanics, C enter for Advanced M echanics and M aterials,T singhua University, Beijing 100084, C hina)A bstract In the past years,the prosperous progress of additive manufacturing,promotes a rapid development of the mechanical architected metamaterials.In this paper,the design and the excellent mechanical properties of these metamaterials are introduced,including the following parts:the evolutional development of the unit cell design from beam-based,plated-based to shell-based structures in order to reduce the intrinsic stress concentration;the route to achieve the isotropic Hashin-Shtrikrnan stiffness limit;their applications in various fields of biological medicine,energy,environment,and others.K ey words mechanical architected metamaterial,ultrahigh strength,isotropic,Hashin-Shtrikrnan stiffness limit在旅行中，我们常常会看到一些精美绝伦的镂空建筑，如高达324米的埃菲尔铁塔、占地25万平方米的鸟巢体育场，令人叹为观止（图1)。

Band-to-band tunneling in a carbon nanotube metal-oxide-semiconductor field-effect transist

Band-to-band tunneling in a carbon nanotube metal-oxide-semiconductor field-effecttransistor is dominated by phonon assistedtunnelingSiyuranga O. Koswatta,†,* Mark S. Lundstrom,† and Dmitri E. Nikonov‡† School of Electrical and Computer Engineering, Purdue University, West Lafayette,Indiana 47906‡ Technology and Manufacturing Group, Intel Corp., SC1-05, Santa Clara, California95052* Corresponding author. E-mail: koswatta@Abstract – Band-to-band tunneling (BTBT) devices have recently gained a lot of interest due to their potential for reducing power dissipation in integrated circuits. We have performed extensive simulations for the BTBT operation of carbon nanotube metal-oxide-semiconductor field-effect transistors (CNT-MOSFETs) using the non-equilibrium Green’s functions formalism for both ballistic and dissipative quantum transport. In comparison with recently reported experimental data (Y. Lu et al, J. Am. Chem. Soc., v.128, p. 3518-3519, 2006), we have obtained strong evidence that BTBT in CNT-MOSFETs is dominated by optical phonon assisted inelastic transport, which can have important implications on the transistor characteristics. It is shown that under large biasing conditions two-phonon scattering may also become important.Device power dissipation has become a major challenge for the continued scaling of integrated circuits.1,2 One important factor contributing to the overall power dissipation is poor sub-threshold (off-state) properties of modern transistors. This has led to large sub-threshold leakage currents, and it has limited the ability to scale the power supply voltage (V DD ), which has been the preferred way to decrease power dissipation in modern integrated circuits. The sub-threshold swing (S ) for conventional transistors, which determines how effectively the transistor can be turned off by changing the gate voltage (I DS vs. V GS ), has a fundamental limit of 2.3/B S k T q =× mV/decade ≈ 60mV/dec at room temperature, where k B is the Boltzmann constant, T the temperature, and q the electron charge. The need to maintain a certain ratio of on-current to off-current limits scaling of the power supply voltage, aggravating heat dissipation problems in modern high-performance circuits.3The aforementioned limitations on the off-state performance of conventional transistors equally apply to those based on carbon nanotubes (CNTs): both Schottky-barrier transistors (SB-CNTFETs) and metal-oxide-semiconductor field-effect transistors (CNT-MOSFETs) with doped source/drain contacts.4,,,567 It has been experimentally observed that carrier transport in CNTs can be nearly ballistic.8,,910 High-performance CNT transistors operating close to the ballistic limit have also been demonstrated.11,,1213 Recently, p-type CNT-MOSFETs with near ideal gate control and sub-threshold operation close to the fundamental limit of S ≈ 60mV/dec at negative gate voltages (V GS ) have been reported.14,15 At positive voltages, these CNT-MOSFETs begin to turn on again with a swing of 40~50mV/dec which is smaller than the conventional limit of60mV/dec. This regime of operation is attributed to band-to-band tunneling (BTBT),14,16 and should be distinguished from the ambipolar operation in SB-CNTFETs where the ambipolar branch is due to opposite type of carrier injection (electrons in the case of p-type FETs) from the drain Schottky contact. 6,17 Specialized device geometries including that based on BTBT have been actively investigated to obtain improved sub-threshold performance in CNT-MOSFET transistors.18,19 In this paper, we present detailed theoretical simulations for the BTBT operation of CNT-MOSFETs using the non-equilibrium Green’s functions (NEGF) formalism20 for both ballistic and dissipative quantum transport. Comparing these results with the recently reported data in Ref. 15, we provide compelling evidence that the BTBT operation of CNT-MOSFETs is strongly affected by phonon-assisted tunneling.The idealized device structure used in this study, shown in Figure 1, is a p-type CNT-MOSFET with wrap-around high-k (HfO2 ~ κ = 25) gate dielectric with thickness t OX = 2nm, doped source/drain regions (N S/D = 0.6/nm) with L SD = 40nm, and an intrinsic channel length of L ch = 30nm. A zigzag (16,0) CNT with diameter ~ 1.2nm and bandgap ~ 0.7eV is considered. The CNT-metal contacts at the ends of source/drain regions are assumed to be Ohmic, with perfect transparency for carrier transport. In experimental devices the Ohmic contacts with electrostatically induced doping is achieved through a back-gated geometry.14,,1521 Although the device structure is idealized, the device parameters are in agreement with those in Ref. 15. Our simulations are performed for p-type CNT-MOSFETs in correspondence with experiments. Due to the symmetry of conduction and valence bands in CNTs,22 the simulated I-V curves for n-type and p-typetransistors are expected to coincide after reversing the polarity of voltages; and indeed we have verified this equivalence.Figure 1. Simulated p-type CNT-MOSFET with doped source/drain regions and wrap-around gate electrode. See text for device parameters.The NEGF calculations for dissipative transport are performed self-consistently with electrostatic simulations.23 We use the nearest neighbor p z tight-binding Hamiltonian () to describe the CNT electronic structure.z p TB H −20,22 The simulations are carried out inthe mode-space, considering transport through the first conduction and valence subbands (E C1 and E V1), respectively.23,24 The difference of energy to the next highest subbands, i.e. |E C2-E C1| = |E V2-E V1|, is about 370meV so the transport through higher subbands can be neglected for typical biasing conditions. The retarded Green’s function for the device under electron-phonon (e-ph) coupling is given by,1()z r p TB S D scat G E EI H −−⎡=−−Σ−Σ−Σ⎣⎤⎦ (1)where are the self-energies for source/drain reservoirs and /S D Σscat Σ is the self-energy for e-ph interaction determined using the self-consistent Born approximation.25,26 All self-energy functions are energy dependent, and the energy indices are suppressed forclarity. The scattering self-energy is related to the in/out-scattering functions (/inout scat Σ) by,22in out scat scat scat scat i ⎡⎤Σ=−Σ+Σ≡−Γ⎣⎦i † (2)where we have neglected the real part for simplicity. The electron/hole correlation functions, , are given by,/()n p G E //n p r in out r G G G =Σ. (3) The in/out scattering functions /in out Σ in Eq. (3) have the contributions from the contact as well as e-ph interaction functions.25,26 The diagonal elements of the correlation functions, , relate to the induced charge density on the CNT surface, , that is used in determining the self-consistent potential, U , from Poisson’s equation in cylindrical coordinates,/(,,)n p G z z E ()ind Q z ()2(),ind D AQ z N N U r z ε+−+−∇=− (4)where, DN + and A N − are the ionized donor and acceptor doping densities, respectively. It should be noted that we use the Neumann boundary condition, 0U ∇≡G , at source/drainends, indicating charge neutrality in the doped reservoirs. The detailed treatment of the self-consistent procedure can be found in Ref. 23.We include one-phonon and two-phonon scattering processes in our simulation. In/out scattering functions in Eq. (2) for one-phonon scattering with optical phonons of energy ω= are given by,()()()(),,1,,,,inn n scat OP OP z z E R N G z z E R N G z z E ωωωωΣ=+++−==(5) ()()()(),,1,,,,out p p scat OP OP z z E R N G z z E R N G z z E ωωωωΣ=+−++== (6)where N ω is the equilibrium Bose-Einstein distribution. The e-ph coupling parameter, R OP , for optical phonon scattering in an (n,0) zigzag CNT is, ()2212OP OP C OP R J M nm ω==, with J 1 = 6eV/Å, |M OP | calculated according to Ref. 27, and m C the mass of carbon atom. For the (16,0) CNT and the longitudinal optical (LO) phonon mode considered in this study, OP ω= = 195meV and R OP = 0.01eV 2, respectively. In Eqs. (5) and (6) the scattering functions are taken to be diagonal, which is the case for local interactions.23,,2526 Two-phonon scattering is included similarly, with the phonon energy replaced by ħω2-ph = 360meV, corresponding to 180meV zone-boundary optical phonons, and the coupling parameter related to that of the one-phonon transition via,28222()(/OP e ph OP tot ph R E R E ω−Γ=Γ±=2), (7) where is the one-phonon broadening and OP Γtot Γ is the total broadening due to scattering and the source/drain reservoirs. Finally, the current through the device from node z to (z+1) in nearest neighbor tight-binding model is determined by,25,2614(,1)(1,,)(1,)(,1,)2z z n n z z p TB p TB qi dE I H z z G z z E H z z G z z E π→+−−⎡⎤=++−+⎣⎦∫=+. (8)We have performed self-consistent NEGF calculations for both ballistic (0) and dissipative () transport using the efficient numerical algorithms reported in Ref. scat Σ=0scat Σ≠29. The simulation results are compared against the data reported in Ref. 15; the relevant experimental I DS -V GS plot is reproduced in Figure 2 for the sake of completeness. In examining Figure 2 we can observe a few important features of the p-type CNT-MOSFET operation: 1) near ideal (S ≈ 60mV/dec) sub-threshold behavior under conventional MOSFET operation for negative V GS biases, 2) larger on-currents for the conventional operation compared to BTBT regime, 3) S ≈ 50mV/dec (< 60mV/dec conventional limit) is observed for BTBT transport at V DS = -0.01V. The BTBT sub-threshold swing, however , degrades with increasing V DS biases . 4) The onset of BTBT occurs at a smaller V GS with increasing V DS , i.e., the onset V GS bias point moves left on Figure 2 with increasing V DS . These experimental device characteristics will be compared against our computational results in order to elucidate the transport mechanisms.Figure 2. Experimental I DS-V GS data at different V DS biases for the p-type CNT-MOSFET; reproduced from Ref. 15.Ref. 16 presents a detailed study of the BTBT operation in a CNT-MOSFET, and here we summarize the transport mechanism responsible for it. Figures 3 (a) and (b) depict the two main mechanisms for BTBT transport in p-type CNT-MOSFETs with sufficiently short channel lengths. For such devices, due to longitudinal confinement inside the channel region, we observe quantized states in the conduction band. As shown in Figure 3(a), at sufficiently large positive gate biases these quantized states align with filled states of the valence band in the drain region, and direct (resonant) tunneling to the source region becomes possible. This alignment has a sharp onset, thus leading to steep sub-threshold slopes in I DS-V GS characteristics.16 This can be further understood by observing the hole Fermi distribution in the source region, noting that carrier conduction is equally explained through hole transport from source to drain. Here, bottom of the hole Fermi distribution in the source is cut off by the valence band edge, and the top is cut off by theconduction band edge in the channel. The result is a cooler current-carrying distribution,which leads to a smaller subthreshold swing. For moderate gate biases, as shown in Figure 3(b), direct tunneling through the conduction band states is prohibited, and the current will be dominated by thermionic emission of holes under the channel barrier in the valence band.Figure 3. BTBT operation in a p-type CNT-MOSFET: (a) For large positive V GS, direct tunneling of electrons from drain to source through the quantized conduction band states in the channel region, (b) For moderately positive V GS direct tunneling is prohibited, but inelastic tunneling is possible in the presence of optical phonons (dashed arrows), (c) NEGF simulation results for energy-position resolved current density spectrum (logarithmic scale) under 195meV LO phonon scattering at V GS= 0.4V, V DS = -0.3V.If we take into consideration high-energy optical phonon scattering, another transport channel is possible via phonon-induced virtual states as shown by the dashed arrows in Figure 3(b). In this case, for BTBT to occur, only the phonon virtual state needs to align with the empty valence band states in the source, and the drain bias should be large enough to allow the quantized conduction band states to be filled by the drain. The existence of the two transport channels is inferred from energy-position resolved current density spectrum in Figure 3(c). Thus, it can be expected that in the presence of optical phonons the onset of BTBT transport will begin at a smaller gate bias, and the device I-V characteristics will be different from those expected in the ballistic case.16Figure 4(a) shows the simulated I DS-V GS characteristics under ballistic transport for the p-type CNT-MOSFET given in Figure 1. It is seen that the sub-threshold swing for regular MOSFET operation (negative V GS for p-type devices) is indeed at the theoretical limit of 60mV/dec, which is not surprising given the wrap-around gate geometry we have employed in our simulations. The on-current for regular operation is also larger than that for the BTBT operation. The overall device currents observed in our simulator, however, are larger than the experimental ones, which can be attributed to contact resistance in the experiment. The simulated sub-threshold swing in the BTBT regime is S≈ 20mV/dec at V DS = -0.01V. Under ballistic transport, the steep slope for BTBT regime does not degrade for large V DS, contrary to the experimental results in Figure 2. More importantly, the onset of BTBT transport moves to more positive gate voltages with increasing V DS, i.e. to the right, as shown in Figure 4(a). This behavior under ballistic transport is well understood and is attributed to the “charge pile-up”, or, “floating body” effect.6,30Asshown in Figure 3(b), under ballistic conditions at large V DS and moderate V GS, the conduction band states are not aligned with the empty source states but are aligned with filled states in the drain. Carriers can tunnel into these states, resulting in a pile-up of electrons. Their potential causes the bands in the channel to move up, thus requiring an even larger positive gate bias before the onset of BTBT. As a result, the onset point in Figure 4(a) moves right to larger gate voltages with increasing V DS in complete disagreement with the experimental observations.Figure 4. Simulated I DS-V GS characteristics for the p-type CNT-MOSFET (black solid line - 60mV/dec swing): (a) ballistic, (b) one-phonon scattering with 195meV LO mode.Figure 4(b) shows the simulated I DS-V GS results for the model p-type CNT-MOSFET in the presence of phonon scattering. The sub-threshold swing at V DS = -0.01V for the steepest section of the BTBT transport is S≈ 35mV/dec, which is poorer compared to the ballistic case.16 It is also observed to further degrade with increasing V DS, in agreement with the experimental observations in Figure 2. Also, the onset of BTBT transport occurs at more negative V GS (moves to the left) with increasing V DS, which is also in agreement with the experimental data. This shift in the onset of BTBT to lower voltages is due to phonon-assisted transport starting at more negative gate biases for larger V DS. It is also important to note that in the presence of optical phonons, the charge pile-up effect discussed earlier is suppressed since the conduction band states are constantly emptied into the source due to inelastic tunneling (see Figure 3(b)). At this point, by comparing the experimental data with ballistic I-V results versus that with optical phonon scattering, it is clear that the former completely fails to explain the observed BTBT features, while the inelastic transport simulation seems to be essential for the correct description. The value for the on-current in BTBT operation (at V GS = 0.6V) with and without scattering, however, is very similar since at large positive gate biases the device current is mainly determine by the direct tunneling component. Thus, it is apparent that the BTBT operation of CNT-MOSFETs is governed by phonon-assisted transport, and dominates the sub-threshold characteristics in this regime. At this point it should be noted that elastic scattering due to acoustic phonons is not expected to affect the BTBT sub-threshold swing which is mainly determined by the inelastic tunneling due to optical phonons. The saturation current in the BTBT regime, however, can be reduced byacoustic phonon back-scattering when such events become allowed by energy conservation. It should also be pointed out that our simulations show slight reduction of the BTBT on-current at large gate biases, V GS > 0.6V. This reduction is, however, smaller than that seen in Fig. 2 at low drain biases.Figure 5. Simulated I DS-V GS with only one-phonon scattering due to 195meV zone center LO mode (solid lines) and, also including two-phonon scattering due to 2 x 180meV zone-boundary mode (dashed lines). Red-star, blue-circle, and green-square curves are at the same voltages as in Figure 4(b), magenta-triangle curves are at V DS = -0.6V.The device operation under large drain biases is discussed next. The solid-triangle curve of Figure 5 shows the I DS-V GS relationship for the model p-type CNT-MOSFET at V DS = -0.6V under 195meV LO phonon scattering. It is seen that, due to the charge pile-up effect at such a high drain bias, the onset of BTBT is pushed to larger positive voltages (right). On the other hand, Figure 2 indicates that the experimentally observedonset of BTBT does not show such effects at large drain biases, and may move further left to smaller gate voltages. Examining Figure 3(b), it is clear that the onset of BTBT can be moved left to smaller gate biases if there are higher energy optical phonons that could allow inelastic tunneling. The highest energy optical phonons available in CNTs, however, are the ~195meV LO mode. On the other hand, Raman scattering experiments on CNTs report strong evidence for multi-phonon mediated processes (overtones).31,32 For example, the G/-band, attributed to two-phonon Raman scattering involving the zone-boundary optical phonons, is observed to have similar intensities compared to the G-band arising from the zone-center LO phonons.31,32 The I DS-V GS results obtained with inclusion of two-phonon scattering are shown by the dashed curves in Figure 5. One can see that at small drain biases, it produces a negligible difference due to Pauli blocking of the two-phonon scattering mechanism. At larger drain biases (exceeding the energy of two phonons), this mechanism becomes effective. For such drain biases the device current near the onset of BTBT is dominated by two-phonon assisted tunneling due to one-phonon process being energetically inactive, and, the relatively strong e-ph coupling for multi-phonon mediated processes in CNTs, a fact observed in Raman experiments.32 In Fig. 5 the onset of BTBT is indeed moved to lower voltages (i.e. moves left) compared to the one-phonon case, and the overall I-V characteristics look even closer to the experimental data. Thus, it can be concluded that at large drain biases multi-phonon assisted inelastic tunneling might also become important for BTBT transport in CNT-MOSFETs.In conclusion, we have performed detailed simulations for BTBT operation of CNT-MOSFETs using both ballistic as well as dissipative quantum transport. By comparing the simulation results with the experimental data, we conclude that the BTBT regime is dominated by optical phonon-assisted inelastic transport. It also appears that under large biasing conditions, multi-phonon scattering may also become important. The strong effect of optical-phonons on BTBT transport should be contrasted with conventional CNT-MOSFET operation, where their influence is found to be marginal up to moderate biases.33,34 It is observed that BTBT operation can indeed produce sub-threshold swings below the conventional limit of 60mV/dec, which makes these devices attractive for low-power applications. The sub-threshold properties, however, are found to severely degrade under typical biasing conditions, and sensitively depend on phonon energies, device geometry, source/drain doping, etc.16Acknowledgment - The authors acknowledge the support of this work by the NASA Institute for Nanoelectronics and Computing (NASA INAC NCC 2-1363) and Intel Corporation. Computational support was provided by the NSF Network for Computational Nanotechnology (NCN). S.O.K thanks fruitful communications with Xinran Wang of Stanford University and the Intel Foundation for PhD Fellowship support.1 International Technology Roadmap for Semiconductors (ITRS), available at .2 M. Horowitz, E. Alon, D. Patil, S. Naffziger, R. Kumar, and K. Bernstein, IEDM Tech. Digest, pp. 11-17, 2005.3 Y. Taur and T. H. Ning, Fundamentals of Modern VLSI Devices, Cambridge University Press, 1998.4 J. Guo, S. Datta, and M. Lundstrom, IEEE Trans. Electron Devices, v. 51, n. 2, pp. 172-177, 2004.5 D. L. John, L. C. Castro, J. Clifford, and D. L. Pulfrey, IEEE Trans. Nanotechnol., v. 2, n. 3, pp. 175-180, 2003.6 J. Knoch, S. Mantl, and J. Appenzeller, Solid-State Elec., v. 49, pp. 73-76, 2005.7 E. Ungersboeck, M. Pourfath, H. Kosina, A. Gehring, B.-H. Cheong, W.-J. Park, and S. Selberherr, IEEE Trans. Nanotechnol., v. 4, n. 5, pp. 533-538, 2005.8 M. Bockrath, D. H. Cobden, P. L. McEuen, N. G. Chopra, A, Zettl, A. Thess, and R.E. Smalley, Science, v. 275, pp. 1922-1925, 1997.9 S. Frank, P. Poncharal, Z. L. Wang, and W. A. de Heer, Science, v. 280, pp. 1744-1746,1998.10 J. Kong et al, Phys. Rev. Lett., v. 87, n. 10, 106801, 2001.11 A. Javey, J. Guo, Q. Wang, M. Lundstrom, and H. Dai, Nature, v. 424, pp. 654-657, 2003.12 A. Javey, J. Guo, D. B. Farmer, Q. Wang, E. Yenilmez, R. G. Gordon, M. Lundstrom, and H. Dai, Nano Lett., v. 4, n. 7, pp. 1319-1322, 2004.13 Y.-M. Lin, J. Appenzeller, Z. Chen, Z.-G. Chen, H.-M. Cheng, and Ph. Avouris, IEEE Electron Device Lett., v. 26, n. 11, pp. 823-825, 2005.14 J. Appenzeller, Y.-M. Lin, J. Knoch, and Ph. Avouris, Phys. Rev. Lett., v. 93, n. 19, 196805, 2004.15 Y. Lu, S. Bangsaruntip, X. Wang, L. Zhang, Y. Nishi, and H. Dai, J. Am. Chem. Soc., v. 128, pp. 3518-1519, 2006.16 S. O. Koswatta, M. S. Lundstrom, M. P. Anantram, and D. E. Nikonov, Appl. Phys. Lett., v. 87, 253107, 2005.17 M. Radosavljevic, S. Heinze, J. Tersoff, and Ph. Avouris, Appl. Phys. Lett., v. 83, n. 12, pp. 2435-2437, 2003.18 S. O. Koswatta, D. E. Nikonov, and M. S. Lundstrom, IEDM Tech. Digest, pp. 518-521, 2005, and references therein.19 J. Appenzeller, Y.-M. Lin, J. Knoch, Z. Chen, and Ph. Avouris, IEEE Trans. Electron Devices, v. 52, n. 12, pp. 2568-2576, 2005.20 S. Datta, Quantum Transport: Atom to Transistor, 2nd ed., Cambridge University Press, Cambridge, MA, 2005.21 A. Javey, J. Guo, D. B. Farmer, Q. Wang, D. Wang, R. G. Gordon, M. Lundstrom, and H. Dai, Nano. Lett., v. 4, n. 3, pp. 447-450, 2004.22 R. Saito, G. Dresselhaus, and M. S. Dresselhaus, Physical Properties of Carbon Nanotubes, Imperial College Press, London, 1998.23 S. O. Koswatta, et al, “Non-equilibrium Green's function treatment of phonon scattering in carbon nanotube transistors,” submitted, e-print: cond-mat/0702496.24 R. Venugopal, Z. Ren, S. Datta, M. S. Lundstrom, and D. Jovanovic, J. Appl. Phys., v. 92, n. 7, pp. 3730-3730, 2002.25 R. Lake, G. Klimeck, R. C. Bowen, and D. Jovanovic, J. Appl. Phys., v. 81, n. 12, pp. 7845-7869, 1997.26 A. Svizhenko and M. P. Anantram, Phys. Rev. B, v. 72, 085430, 2005.27 G. D. Mahan, Phys. Rev. B, v. 71, 205318, 2005.28 P. Y. Yu and M. Cardona, Fundamentals of Semiconductors, Springer, Berlin, Germany, 1996.29 A. Svizhenko, M. P. Anantram, T. R. Govindan, B. Biegel, and R. Venugopal, J. Appl. Phys., v. 91, n. 4, pp. 2343-2354, 2002.30 G. Fiori, G. Iannaccone, and G. Klimeck, IEEE Trans. Electron Devices, v. 53, n. 8, pp. 1782-1788, 2006.31 M. S. Dresselhaus, G. Dresselhaus, R. Saito, and A. Jorio, Physics Reports, v. 409, pp. 47-99, 2005.32 F. Wang, W. Liu, Y. Wu, M. Y. Sfeir, L. Huang, J. Hone, S. O’Brien, L. E. Brus, T.F. Heinz, and Y. R. Shen, Phys. Rev. Lett., v. 98, 047402, 2007.33 J. Guo and M. Lundstrom, Appl. Phys. Lett., v. 86, 193103, 2005.34 S. O. Koswatta, S. Hasan, M. S. Lundstrom, M. P. Anantram, and D. E. Nikonov, Appl. Phys. Lett., v. 89, 023125, 2006.。

2023级新视野及新思路Unit2测试卷试题

Test for Unit 2Part I Listening Comprehension (30%)Section ADirections:In this section, you will hear 10 short conversations. You must read the four choices marked A, B, C and D, and decide which is the best answer.1. What does the woman say about a whole package tour?A. It costs less money.B. It offers more privacy.C. It provides more comfort.D. It helps make new friends.2. What are the two speakers planning to do?A. Send these children to school.B. Help clean the tiny,dirty houses.C. Give these children some money and food.D. Have their daughter get to know such a life.3.What will the woman do?A. Travel alone in safe places in India.B. Follow the Indian culture and customs.C. Wear the same clothes as the local women.D. Choose another place to go instead of India.4.What does the woman feel uncertain about?A. Whether Jane will take care of her son and dog.B. Whether she can go to Jane for some advice.C. Whether it is possible to travel with her kid.D. Whether she can drive a jeep in her travel.5. What are the two speakers talking about?A. The human nature of dreaming the impossible.B. Traveling between the past and the future.C. The possibility of traveling back in time.D. The popularity of time travel stories.6.What is the most probable relationship between the man and woman?A. Colleagues.B. Friends.C. Classmates.D. Roommates.7. What do we learn about Harry?A. He is dishonest.B. He is honest.C. He is as honest as his brother Jack.D. He is as dishonest as his brother Jack.8. What qualities does the man look for in his friends?A. Kindness.B. Friendliness.C. Honesty.D. Sincerity.9. What can we infer from the conversation?A. The woman passed the exam.B. The man passed the exam.C. The man told a lie.D. The woman told a lie.10. What does the man mean?A. He lost a watch in the restroom.B. He found a watch in the restroom.C. He found the owner.D. He has turned it into the Lost and Found. Section BDirections: In this section, you will hear two short passages.After each question you must choose the best answer from the four choices marked A, B, C and D.Passage 111. What kind of society is Australia?A. tolerant.B. strict.C. cruel .D. strange.12. What about the clothing people wear in Australia?A. They have special requirements for their clothing.B. There are strict rules about clothing.C. Clothing is an individual choice.D.they always wear traditional dress.13. When meeting a new person,what questions should be avoided?A. age.B. marital status.C. financial position.D. all of the above.14. What is considered as impolite behavior when meeting someone for the first time?A. shake the person’s right hand with your right hand.B. look at the eyes of the people you’re talking with.C. smile.D. kiss or hug.15. How to build a good relationship with people in Australia?A. Put on a suit.B. Use more polite words.C. Speak loudly.D. Wear make up.Passage 216. What does the woman say about the plants in the Amazon rainforest?A. There are so many plants that it is hard to see the sky.B. Many plants are sick because of the lack of sunlight.C. The plants are more spectacular than she has ever dreamed.D. The plants struggle hard to see the sun or the sky above.17. What does the woman say about the local people?A.They enjoy being so far away from technology.B.They live miles and miles away from each other.C.They never travel more than 15 miles along the river.D.They are separated from the world outside of the jungle.18. Why was the old woman in the village so excited?A.Because Emily showed her a great treasure.B.Because Emily helped her win a big prize.C.Because Emily gave her an empty bottle.D.Because Emily gave her a bottle of wine.19. What has the woman learned from her traveling experience?A.Small and common things should also be valued.B.Local people in the rainforest are easier to please.C.Without modern technology life can be quite different.D.Local people in the rainforest take modern life for granted.20. how did the tourists get around the Amazon rainforest?A.by boat.B.by bus.C.by plane.D.by bike.Section CDirections: In this section, you will hear a passage. You are required to fill in the blanks with the exact words you have just heard.I have never been able to understand people who don’t see the 21. __________ in traveling. The common reason is that traveling is a 22. __________ of time and money. I’ve heard some are 23. __________to travel too far away. I can’t help but feel sorry particularly for those who 24. __________ the experience of seeing a new place as a 100 percent 25. __________one. Telling them stories about unusual encounters doesn’t 26. __________in the expected curiosity but a “Why would you wanna go there!?” It makes me lose 27.__________in humanity. Experience is the best teacher and knowledge is 28. __________. The things traveling can teach you are beautiful because you learn to 29. __________in what you see rather than what you are 30. __________. Part II Banked Cloze (10* 1=10%)Directions: Fill in the blanks by selecting suitable words from the word bank. You may not use any of the words more than once.1. She pointed out with tact and ______ exactly where he had gone wrong.2. They envision an equal society, free of ______ and disease.3. Tango is a ______dance, which brings the dancer together in a way words can’t express.4. The school’s approach must be ______ to that of the parents.5. The concept of ______ development has been defined as profitable.Today, humanistic knowledge continues to 6. ______ the ideal foundation forexploring and understanding the human experience. Learning another language might help to 7. ______ you with great insights as well as gain much 8. ______ of different cultures. Taking a close look at a sculpture might make your think about how an artist’s life 9. ______ his creative decisions. Reading a book from another region of the world might help you 10. ______ about the meaning of democracy. Listening to history courses might give you a clearer picture of what the future will be like.Part III. Reading Comprehension (25%)Directions: Read the following passages and choose the corresponding answer. Passage 1Female employees consistently pay lower airfares than men do for the same flights because they tend to book earlier.We compared the airfare paid by employees in the same position within a company for the same class of travel and used a common statistical technique to account for other factors that might affect differences in airfares. We found that women paid on average $18 less per ticket than their male colleagues. Further investigation allowed us to conclude that this gap is largely explained by the fact that women tended to book earlier than men, 1.8 days on average.We wanted to determine what was causing these gender differences in booking business trips so we tested a variety of possible explanations, such as women choosing to plan ahead or male frequent travelers being inclined to book late. None of these explained away the gender gap, so we applied data collected from surveys that express consumer preferences that play a central role in economic decisions, such as patience and risk avoidance.We found that only the concept of "negative reciprocity"-in which an employee who feels unfairly treated engages in negative behaviors, such as spending their company's money less carefully- explains these differences. The surveys showed men tend to exhibit more of these negative behaviors than women.This isn't to say that all men engage in these behaviors-or that booking relatively late is a sign of abnormal behavior. It only means that the gender gap disappears when we plug in the negativereciprocity variable.Prior research on negative reciprocity among workers found that it can result in lower employee motivation, business performances and workplace morale (士气) and culture.Our results show another way these negative behaviors can manifest themselves, like in airline bookings, and add to evidence that women are less likely to engage in them.Companies spend significant sums of money on business travel. While that $18 difference per ticket may seem small, it adds up. Our analysis suggests early booking by women can translate into savings of $1 million a year for a large multinational company with 20,000 regular travelers.1.What did the author's team conclude about the gender difference in airfares from their further investigation?A.It is largely attributed to women booking earlier than men.B.It is largely explained by women's choosing cheaper flights.C.It is mainly accounted for by male employees' readiness to pay more.D.It is due to the fact that women care more about their company's money.2.What did the researchers want to determine by testing a variety of possible explanations?A.What made male frequent travelers book air tickets late.B.What caused women to plan ahead in booking business trips.C.What motivated women to book cheaper flights.D.What accounted for the gender gap in airfares.3.What happened when the negative reciprocity variable was taken into account?A.Both men and women were found to engage in negative behaviors.B.Neither men nor women viewed booking late as a bad behavior.C.The gender difference in airfare expenses no longer existed.D, The gender gap tended to narrow to a significant degree.4.What did prior research on negative reciprocity among workers find?A.It can do more harm to the workplace than to employees.B.It contributes to the male-female divide in the workplace.C.It proves to be counterproductive in a number of ways.D.It can result in increasing labor-management conflicts.5.What does the author emphasize about their analysis in the last paragraph?A.It can help companies increase their savings significantly.B.It can duly contribute to companies’ business performance.C.It can translate women's booking practice into men's behavior.D.It can enhance large multinational companies' competitiveness.Passage 2Daniel waved at his mother who was watching from the kitchen window. She waved back. He was having fun tossing his new football into the air and catching it again. Suddenly, he saw the neighbor’s gray cat crawl along the top of the wooden fence that divided his backyard from the neighbor’s.“You mean old cat!” Daniel shouted. Daniel didn’t like that cat. He had tried to pet it once and got a big scratch. The cat yawned and then scratched the top of the fence. Tiny pieces of wood floated down to the grass. “Stop wrecking my daddy’s fence!” Daniel demanded. He took the football and threw it with all his might. At the last possible instant, the cat jumped out of the way. The ball flew over the fence, passing right where the cat had been.Then he heard a SMASH! “Oh, no!” he groaned. Daniel crept toward the fence and peeked through a crack. He saw his football lying by the wall of his neighbor’s house. And right beside it was a big blue flowerpot cracked on one side. “Oh, no!” Daniel moaned, “What am I going to do?”Daniel walked around to the gate. “I should ask for permission to go into our neighbor’s yard,” he thought. “But then Mom will find out about the bad thing I’ve done. “He didn’t want that to happen. Instead, he checked the latch. It wasn’t locked. Opening the gate, Daniel walked inside. He ran toward his football and quickly picked it up. Staring at the broken flowerpot, he wondered what to do. “Whatever l do, I’dbetter hurry,” he thought. “l shouldn't be here.”Daniel stuffed the pieces of pottery into the pocket of his sweat shirt. “This is really bad,” he mumbled. Grabbing the flowerpot with both hands, he spun it around until the broken spot faced the wall. He smiled. The bad thing was hard to see. “Now I’ve got to hide the pieces,” he thought. Closing the gate behind him, Daniel searched for a place to hide the broken pieces.“Daniel, where have you been?” Mom asked as she headed up the sidewalk. Daniel wanted to say that he’d been playing in the front yard. He wanted to lie. But that would be another bad thing. He remembered all the bad things he had done: throwing the ball at the cat, smashing the flowerpot, going into the neighbor’s yard without permission. He suddenly realized his fourth bad thing: trying to hide the broken flowerpot.And now Daniel wanted to hide all his bad things with a big lie. No, that would be really bad, he decided. “Mom, would you still love me, even if l told you some bad things I've done?” he asked. His mother smiled and said, “Of course, Daniel!” She knelt beside him so that she could listen better. Daniel took a deep breath and pulled a handful of pottery out of his pocket. Mom could see the broken pieces, but that was okay because telling the truth was easier than telling a lie.6. Before seeing the neighbor’s cat, Daniel was having fun .A. waving at his motherB. tossing his football at the catC. watching his mother from the kitchen windowD. throwing his football into the air and catching it again7. Daniel threw his football at the cat with .A. all his strengthB. all his effortC. all his courageD. all his anger8. What did Daniel do after he broke the neighbor’s flowerpot?A. He quickly ran back home.B. He tried to hide the broken pieces.C. He told his mother what had happened.D. He bought his neighbor a new one.9. How many bad things did Daniel think he had done?A. One.B. Two.C. Three.D. Four.10. Daniel’s mother knelt beside him in order to .A. see his face clearlyB. listen betterC. see the broken piecesD. calm him downPart IV. Translation(20%)Directions: Translate the following paragraphs into English.1.目前的艰难境界，哪能阻抑(obstruct) 我们民族革命的前进。

MLX90620ESF-BAB-000;中文规格书,Datasheet资料

(1) Supply Voltage B = 3V
(3) Package options: A = reserved B = 60° FOV C = reserved D = 40° FOV
Example: MLX90620ESF-BAB-000-TU
Functional diagram
Digital Active Thermopile Array
16x4 IR array
General Description (continued)
The results of the infrared sensor measurements are stored in RAM: • 16-bit result of IR measurement for each individual sensor (64 words) • 16-bit result of PTAT sensor Depending on the application, the external microcontroller can read the different RAM data and, based on the calibration data stored in the EEPROM memory, compensate for difference between sensors to build up a thermal image, or calculate the temperature at each spot of the imaged scene. These constants are accessible by the user microcontroller through the I2C bus and have to be used for external post processing of the thermal data. This post processing includes: • Ta calculation • Pixel offset cancelling • Pixel to pixel sensitivity difference compensation • Object emissivity compensation • Object temperature calculation The result is an image with NETD better than 0.5K at 1Hz refresh rate. The refresh rate of the array is programmable by means of register settings or directly via I2C command. Changes of the refresh rate have a direct impact on the integration time and noise bandwidth (faster refresh rate means higher noise level). The frame rate is programmable in the range 0,5Hz…512Hz and can be changed to achieve the desired trade off between speed and accuracy. The MLX90620 requires a single 3V supply (±0,6V). The customer can choose between 3 operating modes: • Normal. In this mode the device is free running under control of the internal state machine. Depending on the selected refresh rate Fps (Frame per second) the chip is constantly measuring both IR and PTAT and is refreshing the data in the RAM with specified refresh rate; • Step. This mode is foreseen for synchronization with an external micro-controller. The internal state machine is halted. If the command ‘StartMeas’ is received via the I2C bus, a single measurement of all IR and PTAT sensors will be done, then the chip will return in wait state. When in wait state the data in RAM can be read. The MLX90620 is factory calibrated in wide temperature ranges: • -40…85 ˚C for the ambient temperature sensor • -50…300 ˚C for the object temperature. Each pixel of the array measures the average temperature of all objects in its own Field Of View (called nFOV). It is very important for the application designer to understand that the accuracy of the temperature measurement is very sensitive to the thermal equilibrium isothermal conditions (there are no temperature differences across the sensor package). The accuracy of the thermometer can be influenced by temperature differences in the package induced by causes like (among others): Hot electronics behind the sensor, heaters/coolers behind or beside the sensor or by a hot/cold object very close to the sensor that not only heats the sensing element in the thermometer but also the thermometer package. This effect is especially relevant for thermometers with a small FOV as the energy received by the sensor from the object is reduced

galvanic replacement method

galvanic replacement method Galvanic Replacement Method: Unlocking the Secrets of NanotechnologyIntroduction:Nanotechnology is a field that has gained immense popularity and significance in recent years. Its applications span a wide range of industries, including electronics, medicine, energy, and materials science. One of the key techniques used in nanotechnology is the galvanic replacement method. This method has revolutionized the synthesis and fabrication of nanomaterials, leading to breakthroughs in various research areas. In this article, we will delve into the intricacies of the galvanic replacement method, exploring its step-by-step process and its impact on nanoscale science.Step 1: Understanding the BasicsTo comprehend the galvanic replacement method, we first need to understand the fundamentals. This method involves the chemical transformation of metal structures through a redox reaction. It typically occurs between two different metals, where one metal isoxidized, and the other metal is reduced. This transformation process leads to the formation of a new nanoscale structure, exhibiting unique properties that differ from those of the starting materials.Step 2: Preparation of Reactant MetalsThe first step in the galvanic replacement method is acquiring the reactant metals. These metals should have diverse properties to facilitate the redox reaction. For example, one metal could be chosen for its high reactivity towards a particular ion, while the other metal acts as a template to control the shape and size of the resulting nanomaterial. The selection of suitable reactant metals forms the foundation for a successful galvanic replacement process.Step 3: Immersion in Ionic SolutionFollowing the acquisition of reactant metals, they are immersed in a solution containing an appropriate ionic compound. The choice of this compound depends on the desired final nanomaterial and its specific properties. For instance, if one wishes to create a magneticnanomaterial, an iron salt solution may be suitable. The immersion of metals in the ionic solution initiates the redox reaction, leading to the replacement of one metal ions with the other.Step 4: Manipulating Reaction ParametersThe galvanic replacement reaction is influenced by various parameters, including temperature, pH, concentration, and reaction time. Manipulating these parameters provides control over the reaction kinetics and morphology of the resulting nanomaterial. Higher temperatures typically accelerate the reaction, while pH and concentration affect the deposition rate and the distribution of the new metal ions. These parameters need to be carefully optimized to achieve the desired nanomaterial properties.Step 5: Characterization and AnalysisOnce the galvanic replacement reaction is complete, the resulting nanomaterial needs to be characterized and analyzed. Various techniques are employed to examine its structure, composition, and properties. Electron microscopy techniques such as transmission electron microscopy (TEM) and scanning electronmicroscopy (SEM) provide high-resolution images, enabling researchers to observe the nanomaterial's morphology and size.X-ray diffraction (XRD) and energy-dispersive X-ray spectroscopy (EDS) are used to analyze the crystal structure and element composition of the nanomaterial. These characterization techniques shed light on the success of the galvanic replacement method and offer insights into potential applications.Conclusion:The galvanic replacement method is a powerful tool in the realm of nanotechnology. By harnessing the redox reaction between metals, researchers can synthesize nanomaterials with unique properties. Understanding the step-by-step process of this method, from selecting suitable reactant metals to characterizing the resulting nanomaterial, empowers scientists to unlock the secrets of nanoscale science. As advancements in nanotechnology continue to reshape various industries, the galvanic replacement method is poised to play a crucial role in the discovery and development of novel materials and technologies.。

FAULT DETECTION FOR LAMINATED CORE

专利名称：FAULT DETECTION FOR LAMINATED CORE发明人：Vladimir Leonov申请号：US13025274申请日：20110211公开号：US20120206162A1公开日：20120816专利内容由知识产权出版社提供专利附图：摘要：A method and system for evaluating the condition of a laminated core of an electric machine, and including a first yoke wound with a first winding and having a first pair of arms spanning between a first pair of teeth of the laminated core, and a second yoke wound with a second winding and having a second pair of arms spanning between asecond pair of teeth. A power supply provides power to an electrical circuit defined by one of the first and second windings to produce a magnetic flux in the yoke corresponding to the one of the first and second windings and defining an excitation yoke. A monitoring module provides a measurement of a characteristic of an electrical circuit defined by the other one of the first and second windings to identify a fault corresponding to an eddy current between individual laminations in the laminated core.申请人：Vladimir Leonov地址：Oviedo FL US国籍：US更多信息请下载全文后查看。

cct射频微波电容

cct射频微波电容英文回答：CCL (Copper Clad Laminate) in RF and Microwave Capacitors.Capacitors are essential components in RF and microwave circuits, providing energy storage and filtering functions. CCL (Copper Clad Laminate) is a commonly used substrate material for RF and microwave capacitors due to its low loss, high thermal stability, and ease of fabrication.CCL is a composite material consisting of a dielectric layer sandwiched between two copper layers. The dielectric layer is typically made of a ceramic material, such as alumina or zirconia, which provides high insulation resistance and low dielectric loss. The copper layers serve as electrodes for the capacitor.The thickness and dielectric constant of the dielectriclayer determine the capacitance value of the capacitor. Thicker dielectrics result in lower capacitance values, while dielectrics with higher dielectric constants result in higher capacitance values.CCL-based capacitors offer several advantages for RF and microwave applications. They have low insertion loss and high isolation, making them suitable for use in high-frequency circuits. They are also resistant to moisture and temperature variations, ensuring reliable performance in harsh operating environments.Furthermore, CCL-based capacitors can be easily fabricated using standard printed circuit board (PCB) techniques, which makes them cost-effective and suitablefor mass production.中文回答：铜覆层板（CCL）在射频和微波电容中的应用。

电致化学发光传感技术的种类及在食品分析中的研究进展

刘树萍，邢珂，韩博林，等. 电致化学发光传感技术的种类及在食品分析中的研究进展[J]. 食品工业科技，2024，45（1）：325−334.doi: 10.13386/j.issn1002-0306.2022110292LIU Shuping, XING Ke, HAN Bolin, et al. Types of Electrochemiluminescence Sensing Techniques and Research Progress in Food Analysis[J]. Science and Technology of Food Industry, 2024, 45(1): 325−334. (in Chinese with English abstract). doi:10.13386/j.issn1002-0306.2022110292· 专题综述 ·电致化学发光传感技术的种类及在食品分析中的研究进展刘树萍1, *，邢　珂2，韩博林2，关桦楠2,*（1.哈尔滨商业大学旅游烹饪学院，黑龙江哈尔滨 150028；2.哈尔滨商业大学食品工程学院，黑龙江哈尔滨 150028）摘　要：电致化学发光传感技术（Electrochemiluminescence ，ECL ）是一种结合了电化学和光化学的分析方法，因其可控性强、灵敏度高和响应速度快等优势在食品分析领域引起了广泛关注。

ECL 是通过改变发射物或共反应物的浓度，使其信号强度发生变化，从而实现对目标物质的灵敏检测。

首先，该文总结了经典ECL 检测体系及基于新型发射物和共反应物的检测体系，并重点介绍了新型发射物中金属纳米团簇和量子点的最新进展，举例阐述其ECL 传感器的结构和检测原理。

其次，综述了ECL 传感器在食品分析领域中的研究进展。

最后对ECL 传感技术的未来发展趋势进行了展望，为食品中营养成分和污染物的检测提供参考，同时也促进该技术的进一步研究，助力未来食品检测发展。

放射性同位素氧化燃烧仪的维护及故障分析

、纟I仪器使用与维修I-5^--5^--5^--5^--5^--5^--5^--5^--5^--5^-放射性同位素氧化燃烧仪的维护及故障分析李可伦1冯政君2（1.南京大学污染控制与资源化研究国家重点实验室，南京210046#.山西大学黄土高原研究所，太原030006）摘要：放射性同位素氧化燃烧仪是采用氧化燃烧法对放射性14C和3H固体样品进行预处理的仪器&由于仪器高度的集成化和自动化，所以在其日常使用时更需加强维护&基于PerkinElmer Model307型号氧化燃烧仪在国家重点实验室公共平台的使用经验，总结了该仪器的日常维护及故障解决方法，以期最大化该仪器的使用率&关键词：氧化仪日常维护故障分析DOI：10.3969/j.issn.1001—232x.2021.03.035Maintenance and troubleshooting of radioisotope sample oxidizer.Li K4un',Feng Zhengjun2(1.State Key Laboratory of Pollution Control and Resource Reuse,School of the Environment,Nanjing Univers--ty"Nanjing210046,China;2.Institute of Loess Plateau,Shanxi University,Taiyuan030006,China) Abstract:Radioisotope sample oxidizer is a kind of instrument which employs oxidizing combustion method to pretreat14C and3H activities in solid samples separately.The sample oxidizer is highly integratedandautomated,soitneedscarefu l y useand more maintenance.Based on the experience of PerkinElmer model307sample oxidizer in the public platform of national key laboratory,several routine maintenanceandtroubleshootingmethodsoftheinstrumentaresummarizedtomaximizetheusee f iciency.Keywords%Sampleoxidizer#Routinemaintenance#Troubleshooting放射性同位素氧化燃烧仪（以下简称氧化仪）利用氧气和高温条件，将标记有14C或/和3H同位素的土壤、动植物组织、水体等样品转化成14CO2或/和3h2o+1'2]o目前南京大学污染控制与资源化研究国家重点实验室配备有珀金埃尔默公司生产的Model307氧化仪，用于对放射性的环境或生物样品进行预处理&与其它型号的氧化仪不同，该仪器采用非催化剂燃烧的形式，利用自动的燃烧程序,将样品送入燃烧室&整个燃烧的过程不仅避免了催化剂对样品的影响，也更具安全性&此外，双重标记的样品通过该仪器可实现14C和3H两种核素的物理分离，以便后续使用液体闪烁计数仪分别测定两种核素的放射性强度[3\氧化仪的主要使用场所为核电站或分析测试公司⑷，由专人操作使用&而该仪器应用在公共实验平台的情况较少，因为其内部易损部件较多，不同使用者的操作或维护稍有不慎，都会出现故障&为此，本文探讨了氧化仪在公共实验平台的日常维护经验并分析总结了常见故障的解决方法，旨在保证测试结果的准确，进一步提高仪器的使用价值&1氧化仪的工作流程及原理Model307氧化仪由燃烧室、14C和3H收集系统、水氮氧气路系统和废气废液收集系统组成®（如图1所示）氧化仪的燃烧程序由电路和气路共同控制，整个过程高度自动化。

A Dynamic Partitioning Protection Routing Technique in WDM Networks ￠

A Dynamic Partitioning Protection Routing Technique in WDMNetworksAjay Todimala and Byrav RamamurthyDepartment of Computer Science and EngineeringUniversity of Nebraska-LincolnLincoln NE68588-0115U.S.A.Email:ajayt,byrav@Abstract:With the increasing popularity of bandwidth-intensive applications,network trafﬁc has been growing exponentially.Wavelength Division Multiplexing(WDM)is a technique that can harness the huge bandwidth available in an opticalﬁber to satisfy the network demand.Due to the multiplicity of the connections on any single link,fault tolerance is of utmost importance in such WDM networks.Active research on providing fault tolerance in WDM networks in recent times underscores its signiﬁcance.In this study,a dynamic partitioning protection routing technique for routing a Fault Tolerant Path Set(FTPS)in a network is proposed.Wavelength assignment schemes to this FTPS with backup multiplexing technique for networks with no wavelength conversion and full wavelength conversion are discussed.Performance is measured in terms of blocking probability and time to recover the failed connections.Our simulations show that this protection technique performs better than other proposed techniques.Keywords:Optical networks,WDM,fault tolerant touting,fault tolerant path set(FTPS),ﬁxed parti-tioning,dynamic partitioning.1IntroductionAll-optical networks employing WDM technique are fast emerging as the most appropriate choice for next generation networks.Wavelength Division Multiplexing(WDM)is a technique that divides the tremendous bandwidth of an opticalﬁber into many non-overlapping wavelength channels.Due to multiplicity of con-nections on a single link,failure of a single link causes loss of huge bandwidth.Hence fault tolerance carries great signiﬁcance in WDM networks.Fault tolerance can be provided either at the optical layer or by the higher layers(like IP,ATM)[1].The recovery time at higher layers is in the order of seconds.Due to the high data rate of a wavelength channel, the amount of data lost is immense when a connection cannot be restored soon enough.Providing fault tolerance at the optical layer,the connection recovery times can be reduced to the order of milliseconds. Implementing fault tolerance at the optical layer,resources can be more efﬁciently utilized than at higher layers[2].These observations clearly account for the active research on fault tolerance at the optical layer in recent times.Fault tolerant techniques provide routing and wavelength assignment solutions for providing fault toler-ance in WDM networks.In real time not all the links fail at the same time.So we assume single-link failuremodel where only a single-link fails at any given time.Exploiting this,fault tolerant techniques address the RW A problem to improve the blocking performance of the network and at the same time minimize the network-wide average and maximum down-time of the failed connections.Fault tolerant techniques are broadly classiﬁed into,restoration or reactive and protection or pro-active techniques[3,4].In restoration techniques,search for a restoration path is initiated upon occurrence of a link failure.A restoration path is computed and a lightpath is established based on the current network status[3].While restoration methods provide better resource utilization,they suffer from long delays to search and reroute the trafﬁc on to the backup path and there is no guarantee that the connection can be restored upon failure.They provide best-effort type of service.Restoration techniques are classiﬁed into link-based and path-based[3] techniques.Article[4]provides a survey on the restoration schemes available.In protection techniques,when a request arrives it is serviced byﬁrst searching for a physical path. Wavelengths are assigned along the physical path setting up a primary lightpath.When a link fails,the trafﬁc is routed on to another lightpath called a backup lightpath.The resources for backup lightpath/s are reserved before the connection is established on the primary lightpath.A protection method avoids long delay to route trafﬁc onto backup lightpath upon link failure.Hence they provide transparency to higher layers.Protection techniques also ensure that a connection can be restored in the event of failure of a physical link.Transparency and guarantee are provided at the cost of increased blocking probability.In this work we explore protection techniques to improve blocking performance of the network.In the rest of the document a physical path is simply referred as path.A physical path along with the wavelength assignment is referred as lightpath.The rest of the paper is organized as follows.Section 2gives an overview of the classiﬁcation of the fault tolerant protection techniques.Section3discusses the underlying network architecture and the protection model.Section4elaborates computation of the Fault Tolerant Path Set(FTPS)using dynamic sub-path protection routing technique.Section5discusses the different wavelength assignment techniques for the FTPS.Section6presents the experiments and the results.Section7provides the concluding remarks.2Classiﬁcation of Fault Tolerant Protection TechniquesFault tolerant protection or simply protection techniques can be classiﬁed into fault tolerant routing tech-niques and fault tolerant wavelength assignment techniques based on RW A problem in WDM networks. Fault tolerant routing techniques address the problem of routing FTPS for a given node pair in a network. Fault tolerant wavelength assignment techniques address the problem of wavelength assignment to FTPS.2.1Fault tolerant path set(FTPS)Fault tolerant path set is a set of paths capable of providing fault tolerance to a given node pair under single link failure model.A path can be selected from the FTPS to serve as the primary path and the remaining paths can serve as backup path/s.The FTPS for a given node pair can be deﬁned asis many-to-one function)where is set of all possible paths between node pair and is a mapping from to and.The mapping gives the backup path for a given link failure.The fault tolerant path set for a given node pair depends on the routing technique.For the path protection routing technique,for any given node pair is a set of two link-disjoint paths.2.2Classiﬁcation of fault tolerant routing techniquesFault tolerant routing or simply routing techniques can be classiﬁed intoﬁxed partitioning and dynamic partitioning techniques.Inﬁxed partitioning techniques,the primary path is partitioned into sub-paths basedon ﬁxed pre-determined criteria.The ﬁxed partitioning techniques can be classiﬁed into uniform and non-uniform partitioning techniques.In uniform partitioning,the primary path is partitioned into sub-paths of constant length with the exception that the last sub-path can have a length less than .The sub-path length is pre-determined.In a non-uniform partitioning technique,the primary path is partitioned into non-uniform sub-paths based on criteria such as existing network partitions.ProtectionProtection(h = 1)(h = D)(1 < h < D)(1 < h < D)(h = 1)(h = D)Link Protection Figure 1:Classiﬁcation of fault tolerant routing techniques;is the length of the sub-path and is thediameter of the network.Uniform and non-uniform partitioning techniques can be classiﬁed into sub-path and path replacement techniques.In sub-path replacement ,the failed sub-path is replaced by its backup sub-path.With path re-placement ,when a sub-path fails the whole primary path is replaced by its corresponding end-to-end backup path.A uniform partitioning technique with sub-path replacement was proposed in [5].A non-uniform partitioning technique with sub-path replacement is proposed by [6].Figure 1gives a comprehensive clas-siﬁcation of the fault tolerant routing techniques.The uniform partitioning technique with sub-path replacement has two special cases.In the ﬁrst case when the sub-path length equals one it is traditional link protection [5].Path protection is a special case of uniform partitioning technique where a primary path is partitioned into only one sub-path (i.e.,,diameter of the network).It is easy to notice that path protection falls under both sub-path replacement and path replacement techniques as shown in ﬁgure 1.Path protection technique has better resource uti-lization while link protection technique has lower network-wide average down-time [2].The network-wide maximum down-time for path protection does not scale well with the size of the network.When sub-path is not either of the two cases then uniform partitioning technique is called proper sub-path partitioning technique.Figure 2illustrates typical proper sub-path partitioning technique with sub-pathPrimary PathBackup path (p ) for sub-path A-B-C.1Figure2:Typical proper sub-path partitioning technique with sub-path replacement;The sub-paths are A-B-C,C-D-E,and E-F and their length are2,2,1respectively.replacement where sub-path length equals two.The primary path is partitioned into three sub-paths A-B-C,C-D-E,E-F.The last sub-path E-F has length one.(A-M-N-O-C)is the backup sub-path for the sub-path A-B-C of.It is important to note that is link-disjoint with the sub-path A-B-C of.Similarly the backup sub-paths of sub-paths C-D-E and E-F are link-disjoint with their corresponding sub-paths.This technique lowers the average and maximum network-wide down-time but due to theﬁxed partitioning of the primary path irrespective of the network topology and node pair might result in increased blocking.A variation of uniform partitioning technique for node failure is proposed in[7].Dynamic partitioning technique addresses the problem of reducing the network-wide average and max-imum down-time while minimizing blocking performance of the network.In a dynamic partitioning tech-nique,the primary path is partitioned into sub-paths dynamically.The dynamic partitioning techniques can also be classiﬁed based on sub-path replacement and path replacement.With sub-path replacement,the failed sub-path on the primary is replaced by another link-disjoint sub-path.With path replacement,when any of the links on the sub-path fails the whole path is replaced by its corresponding end-to-end backup path.2.3Classiﬁcation of fault tolerant wavelength assignment techniquesFault tolerant wavelength assignment techniques can be classiﬁed into dedicated backup,backup multiplex-ing and primary backup multiplexing.With dedicated backup[2],the backup path is set up at the same time as the primary.The source always routes the trafﬁc on both primary and backup paths.In the event of a failure of the primary path,the destination node switches to the backup path.In dedicated backup there is no sharing of resources by the backup paths.Though dedicated backup has very high network blocking probability the down-time of the connections is least.In single-link failure model,as atmost one link may fail at any time,not all connections require backup at any given time.Backup multiplexing[8]is a technique which allows multiple backup paths to share a wavelength chan-nel as long as no two connections contend for the shared wavelength,upon link failure.Backup multiplexing exploits the fact that not all backup paths need to be activated when a single link fails.It provides guarantee that the failed connections can be restored.In contrast to dedicated backup,there exists delay to conﬁgure cross-connects along the backup route and route the failed trafﬁc.Primary Backup Multiplexing[9]is a technique that not allows one or more backup paths to share a channel but also a primary.When a primary also shares a channel with one or more backup paths,thebackup paths are not available for restoring failed connections.It can be clearly observed that not all failedconnections can be restored upon link failure.Primary backup multiplexing technique is used to achievepre-speciﬁed network restoration guarantees.For example a network with80%restoration guarantee can beprovided.The advantage of such a technique is that with very little decrease in restoration guarantee therecan be a considerable increase in blocking performance.Backup multiplexing can classiﬁed into four basic schemes based on wavelength assignment policy tobackup path.In all the four schemes the primary path is assigned theﬁrstﬁt wavelength.The four differentwavelength assignment schemes for backup paths are First Fit,Last Fit,Most Shared and Least Shared.Intheﬁrstﬁt scheme,the backup paths are also assigned theﬁrstﬁt wavelength as primary path.In the lastﬁtscheme,the backup paths are assigned the lastﬁt wavelength.Both the schemes have their pros and cons.The lastﬁt wavelength assignment scheme to backup paths for networks with full wavelength conversionwas proposed in[10]and later studied by[11].This work more closely examines both the wavelengthassignment schemes.The third wavelength assignment scheme assigns the most shared wavelength to thebackup paths.In case of a conﬂict eitherﬁrstﬁt of lastﬁt can be assigned.If no shared wavelengthexists,againﬁrst or last free wavelength can be assigned.In the fourth scheme,the least shared wavelengthchannel is assigned to the backup paths.In case of a conﬂict eitherﬁrst or lastﬁt is assigned.Again incase no shared wavelength exists eitherﬁrst or last free wavelength can be used.In this work onlyﬁrst twowavelength assignment schemes to the backup paths are evaluated.3Network Architecture and Protection Model3.1Network architectureFigure3illustrates the network architecture.The physical layer consists of nodes interconnected with a pairof uni-directionalﬁbers.Eachﬁber can carry multiple wavelengths.Each node in the network consists ofa crossconnect(XC)and an access station.Every XC and access station has an array ofﬁxed transmittersand receivers for every pair of uni-directionalﬁbers.The number ofﬁxed transmitters/receivers in the arrayis equal to the number of wavelength channels in theﬁber.The XC and access station are connected bymultiple pairs of uni-directionalﬁbers to eliminate any bottle-neck in the performance of the network dueto resource constraints at the access stations.The distance between the station and XC is negligible whencompared to the distance between XCs.It is also assumed that the probability of failure of the link betweena station and a crossconnect is zero.3.2Protection modelLet us assume that a connection is established from node O to node U as inﬁgure4(step3)and the linkbetween the nodes M and L fails.The failure detection time,,is assumed as10s.Once these nodeslearn about the link failure from the loss of the signal,node M propagates an alarm signal upstream viathe signaling channel.Some of the wavelengths onﬁber links are dedicated for carrying this signalinginformation.Node N receives the information after propagation delay.The propagation delay,,is function of length of theﬁber and the speed of light inﬁber.Node N then checks to see if it is capable of switching thetrafﬁc to the corresponding backup path.Once it determines it is incapable,it sends the same signaling information to the upstream node P.When node P receives the information it starts processing and discovers it is capable of switching the trafﬁc to the backup path.Let us assume that this processing time at any node,,is20s.Node P then sends the signaling information to nodes on the backup path that need to be conﬁgured(Q,R,S,T,U).It can be clearly seen that the down-time(DT)of a failure depends on the lengths of the paths the signaling information has to travel.After these nodes receive the information,theyFigure3:Network and Node Architecture.conﬁgure the switches to setup the backup path.Assuming that the switch conﬁguration is done in parallel at any node,the conﬁguration delay at the XC,,is assumed5ms[12].3.2.1Modeling performance parametersThe performance of a protection technique is measured based on its blocking probability(BP)and down-time(DT)of failed connections.The blocking probability of a protection technique is measured as the ratio of blocked calls to the total number of calls.The DT of a failed connection depends on various factors. Effective length of a backup path,,is deﬁned as the number of links of the backup path not present on the corresponding primary path.Effective length of the primary path w.r.t a tagged link,,is deﬁned as thenumber of links the information has to traverse from one of the adjacent nodes of closest to the source to the node from where the failed trafﬁc can be rerouted onto its backup path.Considering the above protection model,the time taken to propagate the failure to the node which is capable of switching the trafﬁc to backup path is.The time taken to conﬁgure the backup path is.The total down-time of the connection due to failure of link,,is where is the backup path corresponding to link.Let be the set of accepted connections during a simulation.The average down-time of an accepted connection,,where is the number of accepted connections.The network-wide maximum down-time of a routing technique,,.4Dynamic Partitioning Technique with Path ReplacementIn dynamic partitioning technique the primary path is partitioned into sub-paths dynamically depending upon current network status and node pair.Inﬁxed partitioning a primary path is partitioned irrespective of current network topology and node pair.If the backup paths for any of partition does not exist,the connection is rejected.In case of dynamic partitioning,the primary path is partitioned based on the availability of the backup paths.The restriction in case ofﬁxed partitioning no longer applies to dynamic partitioning.This allows increase in the blocking performance of the network.Let us compute the FTPS for a given node pair (s,d)using the dynamic partitioning technique with path replacement.4.1The algorithmTheﬁrst path of the for any given node pair is the shortest hop path between node pair.At this point none of the links of are mapped.The remaining paths of the are computed recursively until the resulting mapping is a many-to-one function.To compute the remaining paths,search for an unmapped link on path in someﬁxed order.Link is temporarily removed from the topology and another shortest hop path is computed.Path is added to the set.The mapping is updated to reﬂect the changes in.The process of computing is repeated until the resulting mapping is a function where no unmapped link exists in set.The pseudo code for the computation of using dynamic partitioning technique with path replace-ment is outlined below.Let be an boolean array of size equal to the maximum number of links in the network,representing path,where if link is present on path then=TRUE else=FALSE. Input:Source and destination pair(s,d).Output:Fault tolerant path set and the mapping function.Algorithm:Compute the shortest hop path between node pair(s,d)using Dijkstra’s algorithm.Copy the path array to temporary path array.0;For from tolink identiﬁer of link on.If is TRUE;Temporarily remove the link()from network andcompute shortest hop path using Dijkstra’s algorithm;Compute the effective length of the path,;//Map all the links present on but not on to.,If()do;//Update the mappings of all the previously mapped links on to,//if the effective length of is less than that of the previously mapped path.,If((())do;//Set the mapping of all the unmapped links on not present on to//and remove the mapped links from.,If(());;The so computed depends upon the order in which the links of path are considered.It can be clearly seen from this mapping that any path from can be selected as primary path and the remaining paths can serve as backup path/s.In this study theﬁrst shortest hop path is selected as the primary path. The number of links on the shortest path between any source destination pair is limited by,the diameter of the network.This gives the upper bound on the cardinality of set as.The maximum cardinality of for all node pairs during our simulation on the ARPA network(ﬁgure6)was4.221Assume link OP has failed.Step 3:Step 2:1Backup path p Backup path p Map: PQ, QR, RS, ST, TU ---> p Map: OI, IH, HJ, JM ---> pMap: OP, PN, NM ---> p Step 1: DestinationSourceFigure 4:Computation of FTPS using dynamic partitioning technique with path replacement.4.2ExampleLet us illustrate the computation ofusing the example in ﬁgure 4.The source and destination pair is .Step one:Find the shortest hop path using Dijkstra’s shortest path algorithm .The number of links on path are 5.Step two:Assuming that the link,are only present on butnot on ,map these links to .A mapping from a link to a path says that the path can serve as a backup path for the link.The links are only present on but not on .For these links set the mapping to .There still exists two unmapped links on .In step three:Assuming that the linkandare only present onbut not on ,set the mapping to .The FTPS,and is a many-to-one function.5Wavelength Assignment Techniques5.1Applying backup multiplexing to FTPS computed using dynamic partitioning tech-niqueBackup Multiplexing technique is used to set up primary and backup lightpaths.The following are the important points to consider during wavelength assignment to backup paths.1.Backup lightpaths of a particular connection can share resources among themselves without any con-straint under the single link failure model.2.The backup lightpath of a connection can share resources with its primary.3.The backup lightpath of a connection can share resources with a backup lightpath of any other con-nection as long as there does not exist any common link on their respective primary lightpaths.21Primary path qBackup path q Backup path q Map: LK ---> q Map: QP, PN, NM, ML ---> q 21Figure 5:Illustration of backup multiplexing technique for dynamic partitioning technique.S Figure 6:ARPA physical network topology.Let us illustrate this with an example.Let ,,and ,,be the FTPSs for connections and as shown in ﬁgure 4(step 3)and ﬁgure 5respectively,routed in the same network shown in Figure 6.Let us suppose and are the primary lightpaths for connections and respectively.The remaining paths of the set are the backup lightpaths for their respective connections.As paths and share links ML and LU,can be multiplexed on the same channel as on these links.Link OP is shared by paths and of connection,so can be multiplexed on the same channel on link OP as.Links OP,PN,NM on path are mapped to backup path and link KL on path is mapped to backup path.Since there does not exist any common link,and can be multiplexed on the same channel on links IH and OI.It can be noticed that even though the primary paths of connection and have links in common their backups can still share channels.5.2Network without wavelength conversionIn this section wavelength assignment to FTPS for networks with no wavelength conversion is discussed. Without wavelength conversion all the links on a path should be assigned the same wavelength,wavelength continuity constraint.The wavelength assignment scheme employed for the primary path is First Fit.Let us discuss wavelength assignment scheme to backup paths.For each wavelength index,calculate the cost of routing the backup path over that wavelength.The following is the cost scheme.1.The cost of using a wavelength channel which is already reserved is zero.2.The cost of using a wavelength channel that is used by its own primary path is zero.3.The cost of using a wavelength channel that is not reserved is one.4.The cost of using a wavelength channel that is already used by some other primary is inﬁnity.But while using a wavelength channel that is already reserved for some other backup path/paths,ensure that there does not exist a common link from among the corresponding links of the backup paths already multiplexed on the channel and backup path being considered.In Backup Multiplexing First Fit(BMFF),ﬁrst wavelength index with minimum cost is assigned to the backup path.In contrast,Backup Multiplexing Last Fit(BMLF)scheme selects the last wavelength index with the minimum cost for the backup path.5.3Network with wavelength conversionThe wavelength assignment technique used for primary path is First Fit.For theﬁrst link,loop according to the increasing order of wavelength index toﬁnd a wavelength that is free.Theﬁrst free wavelength is assigned to theﬁrst link.For the subsequent links,check to see if the wavelength on which previous link was routed is free.If it is free,assign it to the link else loop through the increasing order of wavelength index toﬁnd theﬁrst free wavelength.Let us discuss the Backup Multiplexing First Fit(BMFF)and Backup Multiplexing Last Fit(BMLF) wavelength assignment schemes for a backup path.Both the schemes loop through every link on the backup path toﬁnd a wavelength assignment for the link.If is also present on its primary path,then the same wavelength as that of the primary is assigned.The distinction between the two schemes starts when is not present on its primary path.In such a case,loop according to the increasing order of wavelength index toﬁnd a wavelength that is already reserved for some other backup path.But while using a wavelength that is already reserved for other backup path/s,ensure that there does not exist a common link from among the corresponding links of the backup paths already multiplexed on the channel and the backup path in consideration.In BMFF wavelength assignment scheme,ﬁrst such reserved wavelength is assigned.In contrast,BMLF selects the last such reserved wavelength.If such a reserved channel does not exist,the BMFF scheme assigns theﬁrst free wavelength channel that is not reserved.While BMLF assigns the last free wavelength.If no such channel exists,the connection is blocked.However in all cases,the wavelength channel must not be used by any other primary path.6Experiments and ResultsExtensive simulations are performed on the ARPA network topology(ﬁgure6).The ARPA network has21 nodes and52links.The length of each link is assumed as1000km.The calls arriving into the network have Poisson distribution with mean arrival rate.The connections have exponential holding time with mean .The load on the network is measured as.The number of calls simulated in the network duringeach simulation run are100,000.In this work we evaluate four routing techniques and two wavelength assignment techniques.The uni-form partitioning technique with sub-path replacement and sub-path length equals one is traditional link protection(LINK).The uniform partitioning technique with sub-path replacement and sub-path length equals diameter of the network is path protection(PATH).Let the uniform partitioning technique with sub-path replacement and sub-path length()equals two,be referred to as UNIFORM-2.And let the dynamic partitioning technique with path replacement be referred as DYNAMIC.Backup multiplexingﬁrstﬁt wave-length assignment scheme is referred as BMFF and backup multiplexing lastﬁt as BMLF.6.1Evaluating fault tolerant routing techniquesThe performance of the routing techniques LINK,PATH,UNIFORM-2,DYNAMIC are evaluated using BMFF wavelength assignment technique on ARPA network topology.The performance is evaluated in terms of network blocking probability,network-wide average and maximum down-time of failed connections respectively.6.1.1Networks without conversionLet usﬁrst consider network with no wave conversion.Figure7shows the blocking probability(BP)vs.load (in Erlang’s)for PATH and DYNAMIC protection techniques.The blocking probability of DYNAMIC protection technique is lower than PATH protection technique but increases rapidly as the network load increases.The blocking probabilities of the UNIFORM-2and LINK protection techniques are in the or-der of0.4and0.7respectively.Figure8plots the network-wide average down-time()of LINK,PATH, UNIFORM-2,and DYNAMIC protection techniques.The of a protection technique is fairly constant with load.It can be clearly seen that,of DYNAMIC protection technique is lower than that of PATH protec-tion technique.Though UNIFORM-2and LINK protection techniques have lower values,their blocking probabilities are far too high for any fair comparison of their values.The graphs clearly show how dy-namism of the proposed protection technique efﬁciently exploits the network resources when compared to other protection techniques.Figure9plots the network-wide maximum down-time()of LINK,PATH,UNIFORM-2,and DY-NAMIC protection techniques.It can be clearly seen that the network-wide maximum down-time of PATH protection technique is too high when compared to other techniques.It is important to note that the of PATH protection technique does not scale well with the size of the network.As the size of the network increases,the of PATH protection technique increases accordingly.The DYNAMIC protection technique partitions the primary path into multiple sub-paths eventually decreasing the maximum down-time of a con-nection.The DYNAMIC protection technique has as low as that of LINK and UNIFORM-2protection techniques.6.1.2Networks with conversionFigures10,11,12are similar toﬁgures7,8,9respectively except the graphs are plotted for ARPA network with full wavelength conversion.The blocking probability of the DYNAMIC protection technique is lower than PATH protection technique but increases rapidly when the network load increases.It can be seen that。

电致电阻效应研究进展

电致电阻效应文献综述赵重阳2013216020中国石油大学（北京）理学院2013级硕士研究生摘要电致电阻效应因其特殊的物理特性和良好的应用前景引起了人们的广泛关注。

本文主要回顾了电致电阻效应的发现和发展历程，介绍了其产生机理和阻变行为，总结归纳了研究人员对不同材料构成的异质结构中电致电阻效应的研究进展，并且介绍了其应用于下一代非易失性存储技术的研究现状，展望了其良好的应用前景。

关键词：电致电阻；阻态转换；非易失性存储AbstractElectroresistance effect because of its special physical characteristics and excellentapplication prospects has attracted widespread attention. This paper reviews the history of the discovery and development of electroresistance effect, introduces its generation mechanism and resistive switching behavior, summarized the research progress electroresistance effect researchers found in the heterostructures composed of different materials, and introduced its research status for next generation non-volatile memory technology，discussedits excellent prospects.Key words:electroresistance, resistive switching, non-volatile memory1.引言1.1电致电阻效应的发现在简单的金属/金属氧化物/金属三明治结构中，研究发现施加电压能够改变器件的电阻，而且这种改变是持久的且可逆的。

1、下载文档前请自行甄别文档内容的完整性，平台不提供额外的编辑、内容补充、找答案等附加服务。
2、"仅部分预览"的文档,不可在线预览部分如存在完整性等问题,可反馈申请退款(可完整预览的文档不适用该条件!)。
3、如文档侵犯您的权益，请联系客服反馈,我们会尽快为您处理(人工客服工作时间：9:00-18:30)。

A Fault-Tolerant Technique for Nanocomputers:NANDMultiplexingJie Han Pieter JonkerPattern Recognition Group,Department of Applied PhysicsDelft University of TechnologyLorentzweg1,2628CJ,Delft,The Netherlandsjie@ph.tn.tudelft.nl pieter@ph.tn.tudelft.nlKeywords:Fault Tolerant,NAND Multiplexing,Markov Process,Normal Distribution,Binomial DistributionAbstractIn order to make systems based on nanometer-scale devices reliable,the design of fault-tolerant ar-chitectures will be necessary.This paper presents a novel fault-tolerant technique for future nanocomput-ers,NAND multiplexing.Initiated by von Neumann, the NAND multiplexing technique,based on a mas-sive duplication of imperfect devices and randomized imperfect interconnect,had been studied with an ex-treme high degree of redundancy(A4333).In this paper,the NAND multiplexing is extended to rather low degree of redundancy,leading it to a comprehen-sive fault-tolerant theory.The stochastic Markov na-ture in the heart of the system is discovered,and the characteristics of such a Markov chain are exploited. This fault-tolerant technique is potentially useful for future nanoelectronics.1IntroductionThe emerging nanometer-scale devices will proba-bly be less reliable than classical silicon devices for large scale integrated circuits.They are harder to fabricate and more sensitive to external inÀuences such as electromagnetic interference,thermalÀuctua-tions,radiation related effects,etc.Permanent errors may occur during manufacture while transient ones during computers’lifetime.To make systems based on nanometer-scale devices reliable,therefore,fault-tolerant architectures will be necessary.Current fault-tolerant techniques are basically built on redundancy technologies:R-fold Modular Redundancy(RMR)[1](e.g.Triple Modular Redundancy or TMR[2]), Recon¿guration[3][4].With RMR the effect of modest transient errors are effectively eliminated,however some critical compo-nents(e.g.the Majority V oting logic in TMR)have to be highly reliable.Recon¿gurable computers have been successfully implemented in protecting against permanent errors,which are mainly generated dur-ing fabrication,while it would be much less ef¿cient against transient ones[5].As the nanometer-scale de-vices are hopefully much smaller than current CMOS technology,the device failure rate increases due to the limit of manufacturing.In1950s,von Neumann initiated the study of us-ing redundant components to get reliable synthesis from unreliable components,namely,the NAND mul-tiplexing[1].It was then theoretically demonstrated that with an extreme high degree of redundancy,the synthesis of unreliable components could be working reliably.In the following section2and5,von Neu-mann’s theory is brieÀy reviewed and in section3and 4an alternative method is presented and some numer-ical evaluations are performed.In section6we study the error distributions in the NAND multiplexing sys-tem.Finally,there are some discussion and conclu-sions.2von Neumann’s theory on NAND mul-tiplexingConsider a NAND gate.Replace each input of the NAND gate as well as its output by a bundle of Q lines,as shown in Figure1.Duplicate the NAND Q times.The rectangle U is supposed to perform a”ran-dom permutation”of the input signals in the sense that each signal from the¿rst input bundle is randomly paired with a signal from the second input bundle to form the input pair of one of the duplicated NANDs.Let [be the set of lines in the ¿rst input bundle being stimulated (a logic TRUE or %4%) let \be the corresponding set for the second input bundle and let ]be the corresponding set for the output bundle,as -suming the faulty probability of a NAND gate is %and the error behaves as the inversion of the output of the NAND (a von Neumann fault).Let +[>\>],have + { Q> | Q > } Q ,elements,respectively.Clearly + {> |> },are relative levels of excitation of the two in-put bundles and of the output bundle,respectively.The question is then:what is the distribution of the stochastic variable }in terms of the given {and |?With an extremely large Q ,von Neumann’s theory indicates that }is a stochastic variable,approximately normally distributed,with mean +4 { |,.5%+ { | ,and standard deviation s++4 5%, {+4 {, |+4 |,.%+4 %,,@Q [1].3An alternative methodIn this section an alternative method is given to ex-tend the study of the NAND multiplexing technique from extreme high degree to rather low degree of re-dundancy,leading to a comprehensive fault-tolerant theory.Let’s consider a single NAND in the NAND multi-plexing scheme,as shown in Figure 2.We still as-sume that there are {Q and |Q input lines stimu-lated.If the probabilities of error in the two input lines are independent,the probability of any output of the NAND gate that was found stimulated (by at least one non-stimulated input)is }@4 { |(assuming that the NAND gate is fault-free).If each NAND gate has a probability %of making a (von Neumann)error,the probability of its output being stimulated is:}@+4 { |,+4 %,.% { |@+4 { |,.5%+ { |45,(1)Thus,for each single NAND gate,the probability of the output to be stimulated (event 4)is }and the probability to be non-stimulated (event 3)is 4 }.If the Q NAND gates function independently,the en-tire NAND multiplexing unit constitutes a Bernoulli sequence.The distribution of the probability of stim-ulated output lines is therefore the binomial distribu-Figure 1:A NAND multiplexing unit tion.The probability of exactly n outputs being stim-ulated among Q output lines is then:S +n ,@ Qn}+4 },>+n @3>4>5>===Q ,(2)When Q is very large and }is very small,the Pois-son Theorem applies.We have:S +n ,=@olpQn}+4 },@hn $>+Q}@>n @3>4>5>===Q ,(3)Therefore,given Q very large and }very small,the distribution of probability of exactly n outputs from the Q output lines of the NAND multiplexing unit being stimulated,is approximately a Poisson distribu-tion.Empirically the Poisson Theorem gives a pretty good approximation with Q 53, } 3=38=To evaluate the effect of errors,the probability of possible errors below an acceptable level,i.e.S +n {,,needs to be computed.Since the num-ber of the stimulated outputs is a stochastic variable,which comply with the binomial distribution,the De Moivre-Laplace Theorem [6],when Q is very large and 3? }?4,gives:olp Ss{j @]shgw(4)replacing{@ Q}s(5)thenS +n{,(6)=@]ssgwThis shows that the probability of numbers of stim-ulated outputs within the Q output lines of the NANDmultiplexing unit s is approximately normally dis-tributed,with mean Q }@Q ++4 { |,.5%+ { | ,,and standard deviation sQ}+4 },=s++4 5%, { |+4 { |,.%+4 %,,Q ,given Q very large and 3? }?4.Figure 2:An alternative method4Numerical evaluationConsider next the fault distribution of the NAND multiplexing unit for different Q and%within certainranges.We assume that the largest possible error rate %for a future nanoelectronic system is3=4,meaning that one of ten devices is faulty on average.Conse-quently,the%under investigation will be in the range of^3>3=4`.We further assume that the input excitation rates are identical to each other,i.e., {@@ |.This is often true for circuits within the same device.Hence, the fault probability of the output of the NAND mul-tiplexing unit,i.e.the probability of an output line being stimulated,becomes then:}@+4 {,+4 %,.% {@+4 {,.5%+ { 45,(7) For simplicity,we assume that {Q of the Q inputs are not stimulated,so that {@4 {and {?4@5. Replacing {with {in(7):}@+5% 4, {.5+4 5%, {.%(8) For%5^3>3=4`,(8)is monotonically-increasing as {varies from3to4@5.For a typical {,say,3=4, }5 ^3=4<>3=58`.As {increases, }will increase too.This condition does not favor a conclusion in the direction of a Poisson distribution.We proceed with a study on the approximation of a Poisson and a Normal distribution to the binomial dis-tribution for different sizes of the NAND multiplexing unit,i.e.for different Q.We¿rst take Q@4333. Specifying {@3=;and%@43,the probabil-ity(density)of the binomial,Poisson and Normal distribution are plotted in Figure(3).As the prob-ability of possible errors below an acceptable level S+ n {,is an important feature to evaluate the ap-proximations,the cumulative probability distribution S+ n {,for the binomial,Poisson and Normal dis-tribution are plotted as well,in Figure(4).We see that in both¿gures that the Normal distribution is in good accordance with the binomial distribution,while the Poisson distribution is not.The approximation for the Normal distribution is very well kept when {varies in the range^3=:>3=<`and%varies in the range^3>3=4`. Obviously,the larger Q is,the better the approxima-tion.If Q is large enough,the error probability of the NAND multiplexing unit is approximately normally distributed.Now consider the case that Q@433=We still set {@3=;and%@43.The fault probability and cumulative distributions are shown in Figures(5)and (6),respectively.In Figure(5),the probability den-sity of the Normal distribution¿ts in quite well with the samples of the binomial distribution.On the other hand,as the total samples Q is not so large here,the discrete binomial distribution is no longer appropri-ately described by the Normal distribution.Therefore,Figure3:The fault distributions(Q@4333>%@ 43)Figure4:The cumulative fault distributions(Q@ 4333>%@43)Figure5:The fault distributions(Q@433>%@ 43)Figure6:The cumulative fault distributions(Q@ 433>%@43)as shown in Figure(6),neither Normal nor Poisson gives good approximation to the binomial for the cu-mulative distribution.5The restorative unitWe have discussed the setup of a NAND multiplex-ing unit,which executes multiple NAND operations in parallel.We considered the case where the two in-put bundles have almost the same stimulated levels (which is likely in circuits),i.e. {@@ |,with {, |the number of stimulated lines of the two input bun-dles.The error behavior of the NAND multiplexing unit can intuitively be described as follows: If almost all lines of one input bundle are stimu-lated and almost all lines of the other bundle arenon-stimulated,then the probability of the out-put errors(being non stimulated)will approxi-mately be the same as the error probability in ei-ther one of the input bundlesIf almost all lines of both input bundles are non-stimulated,then the probability of the output er-rors(being non stimulated)will be smaller than the error probability in either one of the input bundlesIf almost all lines of both input bundles are stim-ulated,then the probability of the output errors (being stimulated)will be larger than the error probability in either one of the input bundlesWe again assume that {is the error level of the inputs(being non-stimulated),hence, {@4 {. Then the level of the stimulated outputs }is given by equation(8).If%is very small(8)becomes:}@5 { {(9) The(9)is plottedas:When {varies from3to4@5> }is monotonically increasing and } {.This indicates that the non-stimulated inputs give rise to more stimulated out-puts,i.e.the error level is ampli¿ed.If,for example, the original error probability was3=5,the output er-ror probability is3=69.Consequently,we need a unit that restores the original stimulation level without de-stroying the NAND function.Let’s assume a multiplexing system with two types of units.The¿rst being the executive unit which per-forms the NAND function and the second a restorative unit which annuls the degradation caused by the¿rst unit.The restorative unit can be made by using the same NAND multiplexing technique when duplicat-ing the outputs of the executive unit to form the input bundles of the restorative unit,as shown in Figure7a.Figure7a:A restorative unitIf {Q of the Q incoming lines are stimulated and %is very small,according to equation(7),the proba-bility of any output of the restorative unit being stim-ulated }is approximately:}@4 {(10) Plotting }against {=We see that instead of restoring the excitation level, the restorative unit inverts the output of the executive unit,i.e.it transforms the most stimulated bundles to most non-stimulated and vice versa.This suggests that the operation needs to be iterated to construct a proper restoration.Now let the restoring unit consist of two of the restoration units in series,as shown in Figure7.This unit transforms an input excitation level {Q into an output excitation level of(approximately):}@4 +4 {,@5 { {(11) The curve of }against {is shown below,with3 { 4.The curve intersects the diagonal }@ {three times at: {@3> {@3=94;and {@4.If3? {? {, then3? }? { while {? {?4implies that {?}?4.This indicates that the restorative unit brings every {nearer to either0,when {is not larger than 3=94;>or1,when {is not smaller than3=94;.This process has the required restoring effect and hence the unit shown in Figure7gives an effective restoration mechanism.6Error distributions in a multi-stage systemWe have constructed the NAND multiplexing sys-tem of Figure8.In the following parts,we study the error distributions in it.6.1For modest ( fff)We assume that {Q of the Q incoming lines are stimulated for both the inputs of the executive unit in the NAND multiplexing system,and each NANDFigure7:The restorative unit gate has a de¿nite probability%of making a(von Neu-mann)error.Then the probability of any output being stimulated is:}@+4 {,+4 %,.% {@+4 %, +4 5%, {(12)According to the results obtained in section2,the probability distribution of the number of stimulated outputs is the binomial distribution.Hence,the prob-ability of exactly n outputs of the executive unit be-ing stimulated in a bundle of Q lines is given by:S+n,@Qn}+4 },>n5^3>4>5>===Q`(13)The outputs from the executive unit are duplicated and are connected to the successive restore unit in-puts.The same multiplexing as with the executive unit is repeated and the number of stimulated inputs is n5^3>4>5>===Q`.Hence,the stimulated input level {is n@Q.The probability of any output of the ¿rst restorative unit being stimulated is then:}+n,@+4 %, +4 5%, {@+4 %, +4 5%(14)The probability of exactly n outputs of the¿rst restorative stage being stimulated under the condition of n inputs is then:S+n@n,@Qn}+n,+4 }+n,,> n5^3>4>5>===Q`(15) Consequently,the probability of n outputs being stimulated in all cases of n5^3>4>5>===Q`is:S+n,@[S+n@n,S+n,(16)Hereafter,the outputs from the¿rst restorative stage are duplicated and connected to the inputs of the second restorative stage.Now the probability ofFigure8:The NAND multiplexing systemany output being stimulated becomes:}+n ,@+4 %, +4 5%, {@+4 %, +4 5%,+nQ,(17)and the probability of exactly n outputs of the second restorative stage being stimulated under the condition of n inputs is:S +n @n ,@Qn}+n ,+4 }+n ,,>n5^3>4>5>===Q `(18)In all cases the probability of n outputs of the sec -ond restorative stage,i.e.of the complete NAND mul-tiplexing system,being stimulated is then:S +n ,@[S +n @n ,S +n ,(19)@[S +n @n ,+[S +n @n ,S +n ,,In equation (19)the most signi ¿cant parts are the conditional probabilities,S +n @n ,>S +n @n ,and S +n ,(which can be written as S +n @n ,with ¿xed n @ {Q ).For any identical set of inputs and out-puts,all the three conditional probabilities are the bi-nomial distribution with identical parameters,i.e.,S +n@n ,@ Qn}+n ,+4 }+n ,,>(20)n>n5^3>4>5>===Q `where}+n ,@+4 %, +4 5%,+n Q,>n5^3>4>5>===Q `(21)Therefore a +Q .4, +Q .4,matrix ,whose elements are S +n@n ,>n>n 5^3>4>5>===Q `>can be made as (22),so that all conditional probabilities for any set +n>n ,are included.@S +3@3,S +3@4,S +3@5,====S +3@Q ,S +4@3,S +4@4,S +4@5,====S +4@Q ,S +5@3,S +5@4,S +5@5,====S +5@Q ,====================S +Q@3,S +Q@4,S +Q@5,====S +Q@Q ,(22)Accordingly,given a ¿xed input distribution of:S@^s >ss ===s`(23)where s is the probability of l inputs being stimu-lated,the stimulated output distribution in (13),(16)and (19)is given by:S@^S +3,>S +4,>===S +Q ,`@SS@^S +3,>S +4,>===S +Q ,`@SS@^S +3,>S +4,>===S +Q ,`@S(24)The output distribution of the NAND multiplexing system,therefore,can be obtained by (24),given the initial inputs and the conditional probability matrix.6.2A stochastic Markov chainThe number of stimulated outputs of each NAND multiplexing stage is actually a stochastic variable and its state space is D @^3>4>5>===Q 4>Q `.If we name this variable ,where q is the index of the in-dividual multiplexing unit,the evolution of in the NAND multiplexing system is a stochastic process.With ¿xed Q and %,the distribution of for every q is then totally decided by the number of stimulated inputs of the q th multiplexing unit.This can be math-ematically described by:S +5D !@n >@n >===@n,@S +5D !@n,(25)Equation (25)is the condition for a stochastic process to be a Markov process.The evolution of in the NAND multiplexing system,therefore,is a Markov process,or a Markov chain for discrete states and parameters.In a stochastic Markov chain,the transition probability,which indicates the conditional probability from one speci ¿ed state to another,is the most signi ¿cant factor.A transition probability ma-trix like (22)can be made to describe all the transition probabilities.An initial probability distribution and a transition probability matrix are then suf ¿cient to ob-tain all probability distributions in such a stochastic system.We have discussed a NAND multiplexing system with one executive and two restorative stages,wherethe stochastic variables can be described as >and.In principle a system with any number of NAND multiplexing stages,say,q @8>:><>===>can be built to study the fault tolerance in it (note that the odd number is necessary to keep the NAND function).If there is a NAND multiplexing system with q indi-vidual stages in series and its transition probability matrix is given by (22),the output distribution of the multiplexing system is then:S@^S+3,>S+4,>S+5,>===S+Q ,`@S(26)In equation (26)the transition probability matrix for each is identical and irrelevant with regard toq .Thereforeevolves as a homogeneous Markov chain.Because none of the elements in is 3,when q gets large, approaches a constant matrix ,i.e.,olp@(27)and each row of is identical with another.This,as a characteristic of a homogeneous Markov chain,in-dicates that not only the transition probabilities in aNAND multiplexing system will get stable and hardly changes as q becomes extremely large,but also the output distribution will become stable and indepen -dent of the stages of the multiplexing units.6.3is rather large (: fff)If Q is rather large (A 4333),the output error of each NAND multiplexing stage is approximately nor-mally distributed.If there are n stimulated inputs and corresponding n stimulated outputs among Q for each multiplexing stage,the probability density of n is given by:i +n@n ,@ss>3 n Q>3 nQ(28)where}+n ,@+4 %, +4 5%,+n Q,>3 nQ(29)Then the probability of each multiplexing stage hav-ing n stimulated outputs under the condition of n in-puts is approximately:S +n@n ,@i +n@n ,7n>===7n 4(30)The probability of n outputs being stimulated in all cases for 3 n Q is then:S +n ,@[S +n@n ,S +n ,@+[i +n@n ,S +n ,,7n(31)ReplacingS +n ,@i +n ,7n(32)andS +n ,@i +n ,7n(33)we then have in all cases that the probability density of n outputs being stimulated is:i +n ,@[i +n@n ,i +n ,7n(34)In the limit we obtain:i +n ,@]i +n@n ,i +n ,gn(35)The equation (35)is an inductive expression,from which conclusions on the outputs of any NAND mul-tiplexing system can be derived from its initial inputs.We consider that in the multiplexing system of Figure8,for instance,the probability density of the outputs n in terms of the ¿xed inputs n is given by:i +n ,(36)@]i +n @n ,+]i +n @n ,i +n ,gn ,gn@]]i +n @n ,i +n @n ,i +n @n ,gn gnwhere the initial condition is given by:i +n ,@]i +n @n ,i +n ,gn@i +n @n ,(37)In practice the expression (36)is hard to be pre-cisely computed.A practical way is to use the mean of the previous outputs as the ¿xed inputs of the suc-cessive stage,by which the outputs of the system inFigure 8is distributed approximately as:i +n ,@s s>3nQ(38)with}+n ,@+4 %, +4 5%,}+n ,(39)}+n ,@+4 %, +4 5%,}(40)If,for example,a NAND multiplexing system has Q @4333NAND circuits and each NAND circuit has an error probability of %@43,and given that <3(of the initial inputs are stimulated,i.e.,}@3=<,the outputs are approximately normally distributed,with a mean of :4and a standard deviation of ;.Using the normal distribution,a probability of less than 43(of outputs being faulty (stimulated)can be easily computed as being 3=<<<;.If the NAND mul-tiplexing system consists of more stages of restorative units,i.e.,q 8,this probability approaches 4.7DiscussionWe now study the fault tolerance of a NAND mul-tiplexing system with a ¿xed error rate %for each ba-sic logic device,while varying the I/O bundle sizes.It might be interesting to evaluate the performance of a NAND multiplexing system with %@43and with <3(of its inputs stimulated,and the probability that no more than 43(of its outputs is stimulated.The probability distributions versus the number of multi-plexing stages are drawn in Figure 9for different bun-dle sizes Q @43>Q @433and Q @4333.If we pick the number of multiplexing stages to be q @:,then the system has a good performance while the required redundancy (:Q )is not too high.p r o b a b i l i t y o f e r r o r s l e s s t h a n 10%number of multiplexing units (n)Figure 9:Error distribution vs.Number of stages The fault tolerance of this system for different er -ror rates of the NAND circuits can be studied in this speci ¿c case.In Figure 10the probability distribu-tion of errors less than 43(are drawn against the er-ror rate of individual NAND gates,with q @:.It is rather obvious that the NAND multiplexing sys-tem has a better fault tolerance with a larger bundle size Q ,the trade-off,however,has to be made be-tween performance and redundancy.Another conclu-sion is that the NAND multiplexing technique hardly works when the error rate of the basic logic device approaches 3=4=8ConclusionsA fault-tolerant technique based on a massive du-plication of imperfect devices and randomized im-perfect interconnect was comprehensively studied.Within a NAND multiplexing unit with given number of identical NAND logic gates Q ,input error rates {and |,and error rate of the NAND logic %,the prob-ability of numbers of faulty outputs is theoretically a binomial distribution.It can be approximated by a Normal distribution when Q is large (A 4333).The NAND multiplexing system could have more stages of multiplexing units to improve the fault tolerance,when Q is not very large.The error distribution is then a stochastic variable,and evolves as homoge-neous Markov process (chain).The characteristics of such a Markov chain are exploited.Practical simu-lations can be implemented to strengthen these con-clusions.We search for nano-scale basic circuits and shuf Àe mechanisms to implement this technique.9AcknowledgmentWe thank Mike Forshaw of University College London for fruitful discussions.This work is fundedp r o b a b i l i t y o f e r r o r s l e s s t h a n 10%error rate of individual NAND logic (10x)Figure 10:Error distribution vs.Error rate of NAND by Delft University of Technology in its DIRC project ”Novel Computation Structures Based on Quantum Devices”.References[1]J.von Neumann,Probabilistic logics and thesynthesis of reliable organisms from unreliable components.Automata Studies,Shannon C.E.&McCarthy J.,eds.,Princeton University Press,Princeton N.J.43-98(1955).[2]S.Spagocci and T.Fountain,Fault rates innanochip devices,in Electrochemical Society Proc.99-22,354-368(1999).[3]J.R.Heath,P .J.Kuekes,G.S.Snider and R.S.Williams,A Defect-Tolerant Computer Architec-ture:Opportunities for Nanotechnology,Science 280,1716-1721(1998).[4]D.Mange,M.Sipper,A.Stauffer and G.Tem-pesti,Toward Robust Integrated Circuits:The Embryonics Approach.Proc.of The IEEE,88,4,516-541,April 2000.[5]K.Nikolic,A.Sadek,M.Forshaw,Architecturesfor reliable computing with unreliable nanode-vices.Proc.of IEEE-NANO 2001.Page(s):254-259.[6]D.Lu,Stochastic Process and Applications,Ts-inghua University Press,1986.。