Software KK

The software from the division is free to use. The coder should however be acknowledged if the use results in published scientific work.  

https://materialvetenskap.uu.se/research/solid-state-physics+/Software/

KKKTON – A program for Kramers-Kronig (KK) analysis of optical data.

Multifresnel_general – A program for thin film optics calculations. Contains the main program library that is needed for all multifresnel programs listed.

Mulifresnel_sol_vis – Incluces solar and luminous values into the optical calculations.

RT_Invert – A program for calculations of optical constants from experimental data. This program uses the mulitfresnel library.

RT_Invert-variant – A program for calculations of optical constants from the raw data files from the spectrophotometer. This program uses the mulitfresnel library.raw data using the mulitfresnel library.

nk-felyta – Calculates all possible optical constants values from experimental data. Can be used as an aid to determine possible solutions. This program uses the mulitfresnel library.

KKKTON – A PROGRAM FOR KRAMERS-KRONIG (KK) ANALYSIS OF OPTICAL DATA.

The KK analysis is carried out using the extinction coefficient (k) as input and the refractive index (n) is calculated. Low energy Drude and high-energy Lorentz extrapolations are used and parameters for these need also to be inputs. The Matlab program contains a routine for calculating reflectance (R) and transmittance (T) for a film on a substrate and comparing with experimental R and T that need to be entered into the program. kdata.txt is an example of an input file of k-data. See attached files for detailed instructions and program files.

Coder: Annette Hultåker

“MULTIFRESNEL_GENERAL” PROGRAM: THIN FILM OPTICS CALCULATIONS.

This program calculates reflectance, back reflectance and transmittance for a stack of thin films on a substrate. The multilayer stack can have an arbitrary number of layers. A thin film optics formalism using layer and interface matrices is used (P. Pfrommer et al. Sol. Energy, vol 54, No 5, pp. 287-299). The layers can be either coherent or non-coherent, while the substrate is always non-coherent.

Inputs are wavelength (nm), refractive index and extinction coefficient (n and k; from files) for each layer and the substrate, as well as thicknesses (nm), angle of incidence and whether the layers are coherent or not. Wavelength interval and substrate thickness can be adjusted directly in the code.

Output data are plotted in a figure and saved in file ”calcspectra.txt”

Warning: Numerical errors can occur for thick substrates with significant absorption.

The main program calls a number of functions, where different parts of the calculations are carried out. These are:

RT_calc.m  –  Calculates reflectance and transmittance for s-, p- and unpolarized light from the transfer matrix.
transfer_matrix.m  –  Calculates the transfer matrix of the multilayer.
rt2RT.m  –  Calculates reflectance and transmittance from reflection and transmission coefficients.
coherent_l.m  –  Calculates the layer matrix of a coherent layer.
coherent_i.m  –  Calculates the corresponding interface matrix.
non-coherent_L.m  –  Calculates the layer matrix of a non-coherent layer.
non-coherent_I.m  –  Calculates the corresponding interface matrix.
Fresnel_calc.m  –  Calculates Fresnel reflection and transmission coefficients at an interface for s- and p-polarized light.
theta_calc.m  –  Calculates the local angle of incidence at each layer.
mult_3D.m  –  Function for multiplying 3D matrices required for optical calculations.

Note: These functions are also used in all our thin film optics programs on this webpage. By modifying the main program one can use the calculations of the functions for different purposes, for example solving the inverse problem of obtaining optical constants of one layer from experimental data.

MULTIFRESNEL_SOLAR_VIS PROGRAM

This version of the multifresnel program does the same basic calculations as the ”general” version, using the same functions that were supplied with ”multifresnel_general”. However the main program has an additional code that computes solar and luminmous transmittance and reflectance by averaging over the AM1.5 solar spectrum and the luminous eye sensitivity spectrum.

Inputs are wavelength (nm), n and k for each layer, as well as thicknesses (nm) and angle of incidence. The substrate optical constants are also needed. The program needs access, in the same folder, to solar AM1.5 (iso9845tot.txt) and eye sensitivity (eye1.txt) spectra.

Coder: Arne Roos, Anna Werner

RT_INVERT – PROGRAM FOR CALCULATIONS OF OPTICAL CONSTANTS.

These programs fit calculations of reflectance and transmittance for an arbitrary stack of thin films, wherein one layer is unknown, to experimental data. This is done in order to determine the optical constants (refractive index (n) and extinction coefficient (k)) of the unknown layer. Before proceeding to the programs a cautionary note is in order. The inverse problem of thin film optics has multiple solutions and one must be careful to check that the program fits to the physical solution, for example by using independent knowledge of the material under study. Furthermore it is common that the results given by the program cross over from one solution to another one at certain points. This behaviour is especially comon at interference fringes. Hence it may be necessary to perform calculations in different wavelength ranges, in order to obtain the physical solution everywhere. These complications makes determination of optical constants by direct inversion somewhat cumbersome.

The programs listed below all use the MATLAB functions listed in the description of the Multifresnel programs. These must be accessible in the same folder as the main programs listed below.

RT_invert_optconst program

This program calculates optical constants of a coherent or non-coherent layer from an input file containing wavelength, transmittance and reflectance data. The unknown film is situated upon a number of underlayers, which can be either coherent or non-coherent and whose optical constants must be known. The layers are numbered from the top and the substrate is the last underlayer. If number of underlayers is put to zero the remaining layer must be coherent.

Input files are interpolated to the same wavelengths (input data must be given in a wider wavelength region than the interpolation range!). The program uses non-linear least squares fitting starting at the shortest wavelength.

The main program calls the function find_nk as well as RT_calc. The latter one uses the other functions in the multifresnel library.

Input: File with experimental wavelength, transmittance and reflectance. Incidence angle, film thickness and coherence of the unknown layer. Number of underlayers. Optical constants, thickness and coherence of each underlayer.

Output: Figures of n and k; experimental and fitted R and T; deviations between experiment and calculations. Wavelength and calculated n.k are saved in a file.

Coder: Jonas Backholm, Arne Roos (edited by Gunnar Niklasson)

RT_INVERT VARIANT – PROGRAM FOR CALCULATING OPTICAL CONSTANTS

This program is identical to the one above with the following exception: It is intended to be used directly with datafiles from the spectrophotometer. Spectrophotometer measurements start at long wavelengths and proceed to shorter ones. This is often an advantage for the inverse problem, since it is often easier to get good convergence of the calculation at long wavelengths and the problem of multiple solutions is at least less severe. Hence this program uses non-linear least squares fitting starting at the longest wavelength.

The main program calls the function find_nk as well as RT_calc. The latter one uses the other functions in the multifresnel library.

Input: File with experimental wavelength, transmittance and reflectance. Incidence angle, film thickness and coherence of the unknown layer. Number of underlayers. Optical constants, thickness and coherence of each underlayer.

Output: Figures of n and k; experimental and fitted R and T; deviation between experiment and calculations. Wavelength and calculated n.k are saved in a file.

Coder: Jonas Backholm, Arne Roos

NK_FELYTA – CALCULATES ALL POSSIBLE N AND K VALUES. FROM EXPERIMENTAL DATA

This program calculates contours in the n,k-plane pertaining to experimental input values of transmittance and reflectance for each wavelength. Hence the program calculates all n,k values that give those values of R and T and plots them in a figure. The intersections between the R and T contours give all possible solutions of n, k that are compatible with the experimental data. Calculations take long time so it is advisable to use the program for a single or 10-20 wavelengths only.

This type of calculation is especially useful when calculated optical constants fail to reproduce experimental data or when multiple solutions exist and the calculated values cross over from one solution to another. In these cases the contours show which solutions may exist, if for example you are uncertain of whether your solution is the physical one. They also give qualitative guidance on the accuracy of the obtained optical constants. If R and T contours are almost parallel the optical constants will be very uncertain and an intersection may easily be missed due to experimental errors.

The program calls RT_calc and by that all the functions in the Multifresnel library. They are all needed in order to run the program.

Input: An experimental file with transmittance and reflectance data, number of underlayers (substrates), n and k of substrates (Note that the wavelengths must be the same as in the experimental file), thickness and coherence property of the underlayers, angle of incidence.

The output is given as figures appearing on the screen. The video part of the program needs further development. A video movie is played but the figure information is not properly updated in the movie. The video is stored in a file “optcontour” but the frame rate of that movie format seems far too high, so I am not sure if this is useful.

Coder: Jonas Backholm

模拟仿真:太阳能电池研究的利器!

太阳能电池的设计和创新无外乎于两个方面:增强器件的光学吸收性能与增强器件内部的载流子传导收集效率。前者从根本上决定了器件俘获光能的能力,而后者则关乎器件的实际电气输出性能,两者相辅相成,缺一不可。

由于太阳能电池中同时存在光场、电场、载流子分布、边缘界面态等多种相互耦合的物理机制,使得准确预测器件的光电性能变得困难,尤其是在使用了新型的器件构型和光电材料之后。高档次的太阳能电池研究论文往往需要借助计算机模拟仿真来获得器件内部的各种物理信息,为其研究成果提供强有力的理论依据和实验参照。

半导体太阳能电池的计算机仿真技术继承于传统的半导体工艺模拟和器件模拟技术,即TCAD(Technology Computer Aided Design)模拟技术。在传统TCAD软件家族里,最为大家所熟知的两大巨头便是来自Silvaco公司的Silvaco TCAD套装与来自Synopsys公司的Sentaurus TCAD套装。这两款软件包可以实现从半导体器件制造工艺模拟,到分立器件物理特性(电、光、声)仿真,再到电路集成系统性能测试的“全栈式”计算机模拟和设计自动化,因此被广泛使用于现代半导体设计与制造领域,堪称行业标准。其超高的模拟精准度甚至可以用来指导半导体生产线的参数调试。

在光电器件仿真方面,TCAD软件的核心都是先通过各类光学仿真器建立器件内部的稳态光场分布并获得载流子激发速率,再利用有限元分析求解器件内部在指定工作电压下的稳态电场与载流子流场,并最终推算出电极处的光电流强度以及器件的光电能量效率。

传统TCAD仿真软件至今仍然活跃在以硅太阳能电池为代表的传统光伏器件研究领域。例如斯坦福大学崔毅团队便使用了Sentaurus TCAD计算并指导设计了硅基薄膜背接触太阳能电池,相关研究成果被发表在Nature Communications上 (NATURECOMMUNICATIONS | DOI: 10.1038/ncomms3950)。

图 1使用Sentaurus TCAD模拟模拟硅基太阳能电池,获得伏安曲线、内外量子效率等器件特性参数 (NATURE COMMUNICATIONS | DOI: 10.1038/ncomms3950)

随着纳米材料和微纳尺寸器件构型在太阳能电池领域的兴起,光学设计的重要性日益突出。一方面,由于纳米块材,如纳米线、纳米柱等,在空间上天然的稀疏性,或者因为纳米晶薄膜材料中有限的载流子传递效率对材料厚度的巨大限制(通常在几百纳米以内),导致这类器件在光能吸收方面可能有先天的不足。而另一方面,合理使用微纳结构的光学共振特性,能显著提高器件对共振波段的有效吸收,甚至还有可能在光学性能上超越一般的块材器件。

在这样的背景下,传统TCAD软件普遍采用的基于光线追踪(Ray Trace)算法的光学仿真器变得不再适用,与此同时一类更加注重器件微纳光学性能计算,精简载流子输运模拟的仿真模式开始在太阳能电池领域兴起。例如Lumerical公司的FDTD(Finite-differencetime-domain)光学仿真器便是其中的代表。得益于对麦克斯韦方程的直接(数值)求解,这类仿真模式能更加准确地还原器件的各类光学模式和载流子激发分布,尤其在对拥有光子晶体、表面等离子激元等光学现象的器件上有突出的表现。

下面介绍的这一篇来自美国德州奥斯汀分校Shaochen Chen团队的Nano Letters(Nano Lett |DOI: 10.1021/nl904057p)便是一篇时间相对较早的代表性论文。在这篇文章中,研究人员使用了FDTD光学仿真探究了使用金属条栅结构在薄膜电池中获得广谱、广角、偏振不敏感的光吸收增强的可能。其背后的物理机理便是充分利用了薄膜器件中Fabry – Perot共振、平面波导以及金属条栅的表面等离子激元等多种光学模式。而FDTD模拟成为了揭示这一作用机理的利器。

图 2使用FDTD光学模拟准确获取薄膜太阳能电池中的光学模式(Nano Lett |DOI: 10.1021/nl904057p)

在实验方面,香港科技大学范志勇教授团队使用纳米拓印技术将类似上述的器件设计理念应用于超薄非晶硅电池的设计创新中,成功使器件的光学性能获得了约30%的提升,相关研究成果被刊发在顶级神刊物EES上(Energy Environ. Sci| DOI: 10.1039/c3ee41139g)。这里,FDTD光学模拟同样在器件设计指导与论文理论说明上提供了强有力的支撑。

图 3使用FDTD光学模拟还原使用复杂器件构型下的内部光场信息 (Energy Environ.Sci| DOI: 10.1039/c3ee41139g)

对于基于纳米线\柱构型的太阳能电池而言,准确计算预测纳米线\柱结构中的光学共振模式,如Mie共振模式和波导模式,是获得优质光学吸收性能的一大前提。

在这篇曾经引领一时风潮的InP纳米线太阳能电池的报道中(Science|DOI: 10.1126/science.1230969),瑞典兰德大学MagnusT. Borgström教授团队就使用了散射矩阵(scattering matrix)法计算了InP纳米线阵列中的光场分布以及每根纳米线内部相应的载流子激发率。这些模拟结果成为其论证纳米线太阳能电池光学性能的重要理论依据。

图 4  使用光学模拟技术获取InP纳米线阵列中的载流子激发率分布(Science|DOI: 10.1126/science.1230969)

而后来更为登峰造极的经典之作当属下面这篇来自丹麦哥本哈根大学Anna Fontcuberta i Morral教授团队的NaturePhotonics(NATURE PHOTONICS|DOI: 10.1038/NPHOTON.2013.32)。在这篇文章中,研究团队使用了一根看似平淡无奇的p-i-n核-壳(core-shell)纳米单线,通过利用纳米线的Mie共振模式,获得了对太阳光谱有针对性的增强吸收,最终在标准光谱下(AM1.5)实现了超越Shockley–Queisser极限的器件转换效率!助其完成光学理论验证与说明的正是更为精准和完备的FDTD光学仿真。

图 5使用FDTD光学仿真设计单纳米线太阳能电池,获得超越Shockley–Queisser极限的器件转换效率(NATURE PHOTONICS|DOI: 10.1038/NPHOTON.2013.32)

在以量子点薄膜为代表的新兴薄膜材料电池领域,计算机仿真同样在增强薄膜的整体光学吸收性能和载流子输运效率方面显现出强大的指导能力和实用价值。

以量子点电池闻名于世界的多伦多大学Sargent教授课题组自2014年以来,多次在Nano Letters, ACS Nano等顶级期刊上发表有关利用量子点薄膜整体构型来提升其光学吸收性能的研究成果,而其背后同样使用了大量的FDTD光学仿真来论证支持他们的设计。

(NanoLetters|DOI: 10.1021/nl504086v,

ACSNano|DOI: 10.1021/acsnano.5b01296,

NanoLetters|DOI: 10.1021/acs.nanolett.6b05241)

图 6利用FDTD仿真,计算复杂量子点薄膜构型的光学吸收性能(Nano Letters|DOI:10.1021/nl504086v)

而在Sargent课题组更早期一点的研究中(Advance Material| DOI:10.1002/adma.201104832),他们更是借助了Sentaurus TCAD器件仿真来研究纳米柱构型的电极结构在量子点薄膜中收集载流子的能力,成为当时研究纳米晶薄膜电池载流子输运效率的经典之所。

图 7使用Sentaurus TCAD研究了纳米柱构型的电极结构在量子点薄膜中收集载流子的能力(Advance Material| DOI:10.1002/adma.201104832)

总结一下,太阳能电池的器件仿真可以帮助我们获得器件内部准确的光场与载流子激发分布,并可以进一步计算出器件内部的载流子收集传输情况,获得器件的完整伏安工作特性。以此为基础,太阳能电池的四大基础参数,即短路电流、开路电压、填充因子和能量效率,都可以计算得出。在光学仿方面,微纳尺寸构型的太阳能电池应该充分考虑其中的光学共振模式所带来的影响。借助器件模拟技术,可以为器件的形貌设计和材料选择提供坚实的理论指导。

通过上面介绍过的例子我们看到,半导体器件模拟仿真在太阳能电池研究领域的巨大作用。优质的器件仿真可以显著提高研究成果的理论完备性。

电场调控下的浮动薄膜多彩结构

https://mp.weixin.qq.com/s?__biz=MzU5NjM5Njc2Mw==&mid=2247513206&idx=1&sn=a30ab380f95b359ad7223b2f765dd9e9&chksm=fe61abc2c91622d4bd1695b8b41a8e161dcac62f2265667b0a73c807ce1c0aa6892abc1d475f&mpshare=1&&srcid=0425bBu1qLOd4Orj1RKZEElv&sharer_sharetime=1619360087995&sharer_shareid=cfcd208495d565ef66e7dff9f98764da&scene=2&subscene=1&clicktime=1619413444&enterid=1619413444#rd

电场调控下的浮动薄膜多彩结构

Original 长光所Light中心 中国光学Yesterday收录于话题#纳米光子学2#忆阻器1

Image
Image
Image
Image
Image
Image
Image

Banner | 商务推广


中国光学

中国光学中科院长春光机所Light学术出版中心旗下学术传播与服务平台,带你解读全球光学最新研究进展。442篇原创内容Official Account

说明 | 本文是由论文作者投稿
颜色,作为最直观的物理特性之一,在人类文明发展的历史长河中扮演着重要角色。《孟子・告子上》中所写,“目之于色,有同美焉”,更有《道德经》对五色的描述“五色令人目盲”。
对光和色彩的调控是一个亘古不变的问题,我们在惊叹千年前敦煌莫高窟的绚烂壁画的色彩技术时,更致力于对色彩和光的各种研究和应用。
色彩显示技术在材料科学,结构工程,物理,化学,光学和半导体等工业部门都拥有极大的关注,如仿生结构,光子晶体,等离激元,功能材料,3D 打印,全息投影显示,彩色显示器和信息安全等领域。其中,色彩调控技术是各项技术前进和突破的核心指标之一。

Image

图1 敦煌莫高窟彩色壁画
当前针对色彩调控技术的研究,主要围绕在对材料性质的调控,包括化学反应,相变转换,光敏,机械和热响应等功能材料。例如N.Liu等人基于金属等离子基元设计的超高彩色显示器,通过暴露在特殊气氛环境(H₂/O₂)下的氧化还原反应,可以实现色彩的动态调控(代表论文:Nat Commun 8, 14606 (2017).  | 原文阅读 >)。Harish等人巧妙利用相变材料构筑的薄膜结构,非晶态到晶态的转换可帮助器件实现的颜色显著改变(代表论文:Nature 511, 206–211 (2014). | 原文阅读 >)。然而,因为切换前后两种功能已经固定,原位上进行每个单元的任意切换仍然无法得到实现。总结之前的研究,固态器件的物理结构,在加工制备后通常是固定的,而物理结构是光学性能的基础。因此,如何实现固态器件物理结构的调控仍是一个颇具挑战的科学和工程问题。
针对这一个问题,Anders等科研人员在等离子体结构色领域,提出了利用激光调控固态结构的方法:激光热效应可以改变原有的固体结构单元的几何形状,从而实现颜色的改变(Nature Nanotech 11, 325–329 (2016). | 原文阅读 >)。这种技术仍依赖前期复杂的结构制备工艺,而且很难实现精准和可逆化驱动。因此,如何实现简化制备工艺和调控加工成型后固态器件的物理结构,依然是一个巨大的挑战。
薄膜结构因拥有相对简易的制备工艺和高兼容的集成特性,是大部分光学,电学,光电器件,磁性存储,太阳能等器件的基础单元。基于不同材料的多层薄膜结构所显示的光学特性,展示了丰富的颜色及其广泛的应用场景。固体薄膜器件在加工制作完成后,其物理结构通常被固定。
近期,来自新加坡科技大学的Chong Tow Chong教授,联合新加坡国立大学的仇成伟(Qiu Cheng-Wei)教授和清华大学的赵蓉(Zhao Rong)教授,提出了一种可调控的动态薄膜结构设计,能够实现在电场调控的可逆结构色。
研究成果以“Floating Solid-state Thin Films with Dynamic Structural Colour”为题,发表在NATURE NANOTECHNOLOGY上。
该项研究利用了电场下金属Ag离子的迁移活性,巧妙地设计了无序态(amorphous | 名词解释 >)氧化铁为中间介电质层的可调薄膜结构(如图2)。电场驱动下,Ag离子的迁移可以实现薄膜器件的厚度,层数和层序的改变,从而实现单一器件多种颜色的可逆调控。在反向电场下,Ag离子朝向底电极(TiW)方向移动,薄膜器件基本可以恢复到初始状态。研究人员,通过透射电子显微镜(Transmission Electron Microscope)和能量散射X射线谱(Energy-dispersive X-ray spectroscopy)元素表征(图2所示),可以确认该固态器件的物理结构实现了浮动改变。

Image

图2 动态薄膜结构设计:电场调控下的结构变化.图源:Nature Nanotechnology. (2021) Fig.1 (c)-(f)
基于这个动态薄膜的设计,研究人员使用原子力显微镜(Conductive Atomic Force Microscope | 原理解释 >)展示了这项技术应用于超高分辨率的反射型显示器的可能性。如图3,通过导电原子探针的扫描,在一个薄膜样品上,可以写出明显的彩色图形。这些彩色图形,处于non-volatile状态,无需持续的能量去维持色彩区域,这与当前的主流色彩显示技术有很大区别,可在日光下实现清晰的显示(而不需要借助外界的能量的支援)。这项技术将大大降低功耗,为实现节能环保型的新型显示器的研究提供了新的思路。图3 原子力显微镜驱动下的纳米分辨率反射型显示器展示图源:Nature Nanotechnology. (2021) Fig.3(a), (c).
随着纳米加工技术的进步,固态彩色打印技术主要是基于各种纳米等离基元结构的设计。尽管实现了超高纳米级别的分辨率,然而其耗时复杂的工艺对于宏观大尺寸的彩色打印技术,仍然是一个难题。在该项研究中,作者基于图1动态薄膜的设计,尝试了大尺寸的彩色图形的打印,图4展示了在4 inch 硅片上的彩色鱼尾狮图形和‘NUS’(National University of Singapore缩写),证明了该项研究在大尺寸固态彩色打印技术领域的可行性。图4 大尺寸彩色图形打印展示图源:Nature Nanotechnology. (2021) Fig.4(c), (d), (e).
本工作通过电场控制Ag在多层的薄膜结构的迁移,成功驱动了固态薄膜物理结构的浮动变化,进而实现单一器件的颜色大幅度的可逆调控色彩的显示无需持续的能量维持。作者基于此项设计,进一步探索了纳米级别的彩色显示器和大尺寸彩色打印的应用价值。该项研究同时为功能性光学,电学和光电等半导体相关技术的研究提供了新的窗口。
文章信息Yan, Z., Zhang, Z., Wu, W. et al. Floating solid-state thin films with dynamic structural colour. Nat. Nanotechnol. (2021). 
论文地址https://doi.org/10.1038/s41565-021-00883-7直通原文 >编辑 | 赵阳
 欢迎课题组投递成果宣传稿

转载/合作/课题组投稿,请加微信:447882024

往期推荐

Image

Light: Science & Applications | 光的“降噪”:抑制光的反向散射

Image

Nature Photonics | “机器学习”在超快光子学中的应用

Image

Science | 微型光谱仪的30年进展

新刊推荐

Image

开 放 投 稿:Light: Advanced Manufacturing

ISSN 2689-9620

期 刊 网 站:www.light-am.com

开 放 投 稿:eLight

CN 22-1427/O4   ISSN 2662-8643

期 刊 网 站:elight.springeropen.com

点这里 👇 关注我

中国光学

中国光学中科院长春光机所Light学术出版中心旗下学术传播与服务平台,带你解读全球光学最新研究进展。442篇原创内容Official Account

别忘了点赞+在看哦!喜欢此内容的人还喜欢Adv. Electron. Mater.:MXene-GaN范德华异质结,用于高速自驱动光电探测器和发光二极管Adv. Electron. Mater.:MXene-GaN范德华异质结,用于高速自驱动光电探测器和发光二极管…低维 昂维不喜欢不看的原因OK内容质量低 不看此公众号Laser Photonics Rev.:原子薄电致发光器件中的激子发射Laser Photonics Rev.:原子薄电致发光器件中的激子发射…低维 昂维不喜欢不看的原因OK内容质量低 不看此公众号Adv. Funct. Mater.:用于大面积光电探测的2D材料增强型柔性自修复光电探测器Adv. Funct. Mater.:用于大面积光电探测的2D材料增强型柔性自修复光电探测器…低维 昂维不喜欢不看的原因OK内容质量低 不看此公众号

Guide to IBM Cloud

eachers / parents / code club-leaders with unmanaged class accounts need to provide ‘API Keys‘ from IBM Cloud. These are secret codes that provide access to the artificial intelligence technology that will power student projects.Text projects need a ‘Watson Assistant’ API Key. (Images, numbers and sound projects don’t require API Keys).You will need to create an account on IBM Cloud to be able to create these codes. It is free to create an account, and there are no charges for ‘Lite’ API Keys.

How to create API KeysDownload a step-by-step guide for how to create API Keys on IBM Cloud

ML: Keras multiple input multiple output

https://colab.research.google.com/drive/1ZNLC7QMG0Y2Zf4evf6JG_L0rybzmscoj#scrollTo=lwm64LEEt5NX

Models with multiple inputs and outputs

The functional API makes it easy to manipulate multiple inputs and outputs. This cannot be handled with the Sequential API.

For example, if you’re building a system for ranking customer issue tickets by priority and routing them to the correct department, then the model will have three inputs:

  • the title of the ticket (text input),
  • the text body of the ticket (text input), and
  • any tags added by the user (categorical input)

This model will have two outputs:

  • the priority score between 0 and 1 (scalar sigmoid output), and
  • the department that should handle the ticket (softmax output over the set of departments).

num_tags = 12  # Number of unique issue tagsnum_words = 10000  # Size of vocabulary obtained when preprocessing text datanum_departments = 4  # Number of departments for predictions
title_input = keras.Input(    shape=(None,), name=”title”)  # Variable-length sequence of intsbody_input = keras.Input(shape=(None,), name=”body”)  # Variable-length sequence of intstags_input = keras.Input(    shape=(num_tags,), name=”tags”)  # Binary vectors of size `num_tags`
# Embed each word in the title into a 64-dimensional vectortitle_features = layers.Embedding(num_words, 64)(title_input)# Embed each word in the text into a 64-dimensional vectorbody_features = layers.Embedding(num_words, 64)(body_input)
# Reduce sequence of embedded words in the title into a single 128-dimensional vectortitle_features = layers.LSTM(128)(title_features)# Reduce sequence of embedded words in the body into a single 32-dimensional vectorbody_features = layers.LSTM(32)(body_features)
# Merge all available features into a single large vector via concatenationx = layers.concatenate([title_features, body_features, tags_input])
# Stick a logistic regression for priority prediction on top of the featurespriority_pred = layers.Dense(1, name=”priority”)(x)# Stick a department classifier on top of the featuresdepartment_pred = layers.Dense(num_departments, name=”department”)(x)
# Instantiate an end-to-end model predicting both priority and departmentmodel = keras.Model(    inputs=[title_input, body_input, tags_input],    outputs=[priority_pred, department_pred],)

You can build this model in a few lines with the functional API:

Metamaterial Edge and Bandpass Filters

https://www.jobs.ac.uk/job/CDG262/metamaterial-edge-and-bandpass-filters

Bangor University

Metamaterial Edge and Bandpass Filters

Bangor University – School of Computer Science and Electronic Engineering

Location:Bangor
Salary:£11,586 annual stipend
Hours:Full Time
Contract Type:Fixed-Term/Contract
Placed On:18th December 2020
Closes:11th January 2021
Job Ref:BUK2217

This project will be based in the School of Computer Science and Electronic Engineering, in collaboration with Qioptiq Ltd.

Project ID:  BUK2217

Annual Stipend: £11,586

Application Deadline: 11/01/2021

Interviews: 14/01/2021

We are seeking a highly capable and motivated graduate for this exciting opportunity to undertake a Research Masters. This is a collaborative project between Bangor University and leading photonics company Qioptiq Ltd.

This project aims at developing next-generation filters based on metamaterial design and technology, i.e. artificially engineered material with unique properties not found in nature materials. The project will focus on the development of edge and bandpass filters which are building elements of many optical products and systems, whose main function is to limit stray light found in optical systems, especially in underwater environments.

This Research Masters project is expected to start in January 2021 immediately after the interview and will take one year to complete. The KESS 2 scholar will be based in the School of Computer Science and Electronic Engineering (CSEE) and will be supervised by academic supervisors Dr James Wang and Dr Liyang Yue, and the industrial supervisor Dr. James Monks of Qioptiq. The applicant should hold a good degree (at least 2:2) in engineering, physics, optics, materials or a related scientific discipline, have demonstrably excellent research skills, and relevant experience.

For more information, or informal enquiries, please email Dr James Wang : [email protected]

To apply, please send your CV and a cover letter to [email protected] and cc to Penny Dowdney [email protected]

Knowledge Economy Skills Scholarships (KESS 2) is a pan-Wales higher level skills initiative led by Bangor University on behalf of the HE sector in Wales. It is part funded by the Welsh Government’s European Social Fund (ESF) convergence programme for West Wales and the Valleys.

Due to ESF funding, eligibility restrictions apply to this scholarship. To be eligible, the successful candidate will need to be resident in the Convergence Area of Wales on University registration, and must have the right to work in the region on qualification. 

www.kess2.ac.uk