Intelligent Nanotechnology: Merging

Nanoscience and Artificial Intelligence

Yuebing Zheng
Visit to download the full and correct content document:
https://ebookmass.com/product/intelligent-nanotechnology-merging-nanoscience-and
-artificial-intelligence-yuebing-zheng/
INTELLIGENT
NANOTECHNOLOGY
Materials Today: Connecting the
Materials Community
INTELLIGENT
NANOTECHNOLOGY
Merging Nanoscience
and Artificial Intelligence

Edited by

YUEBING ZHENG
Walker Department of Mechanical Engineering and Texas
Materials Institute, The University of Texas at Austin,
Austin, TX, United States
ZILONG WU
Walker Department of Mechanical Engineering and Texas
Materials Institute, The University of Texas at Austin,
Austin, TX, United States
Elsevier
Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands
The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom
50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States
Copyright © 2023 Elsevier Inc. All rights reserved.
No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical,
including photocopying, recording, or any information storage and retrieval system, without permission in writing
from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies
and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency,
can be found at our website: www.elsevier.com/permissions.
This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as
may be noted herein).

Notices
Knowledge and best practice in this field are constantly changing. As new research and experience broaden our
understanding, changes in research methods, professional practices, or medical treatment may become necessary.
Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any
information, methods, compounds, or experiments described herein. In using such information or methods they
should be mindful of their own safety and the safety of others, including parties for whom they have a professional
responsibility.
To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for
any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from
any use or operation of any methods, products, instructions, or ideas contained in the material herein.

ISBN: 978-0-323-85796-3

For information on all Elsevier publications

visit our website at https://www.elsevier.com/books-and-journals

Publisher: Matthew Deans

Acquisitions Editor: Stephen Jones
Editorial Project Manager: Clodagh Holland-Borosh
Production Project Manager: Kamesh R
Cover Designer: Mark Rogers
Typeset by STRAIVE, India
Contributors

Trevon Badloe
Department of Mechanical Engineering, Pohang University of Science and Technology
(POSTECH), Pohang, Republic of Korea
Yijun Bao
Department of Biomedical Engineering, Duke University, Durham, NC, United States
Keith A. Brown
Department of Mechanical Engineering; Physics Department and Division of Materials Science &
Engineering, Boston University, Boston, MA, United States
Qian Chen
Department of Materials Science and Engineering; Department of Chemistry; Beckman Institute
for Advanced Science and Technology; Materials Research Laboratory, University of Illinois at
Urbana-Champaign, Urbana, IL, United States
Hongrui Cheng
College of Chemistry, Fuzhou University, Fuzhou; Fujian Science & Technology Innovation
Laboratory for Optoelectronic Information, China
Liu Chen-Xu
Beijing Jiaotong University, School of Civil Engineering, Beijing, China
Frank Cichos
Leipzig University, Molecular Nanophotonics Group, Leipzig, Germany
Zhengyang Duan
Department of Electronic Engineering, Tsinghua University, Beijing, China
Zheyu Fang
School of Physics, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing,
China
Paul S. Francis
School of Life and Environmental Sciences, Faculty of Science, Engineering and Built
Environment, Deakin University, Waurn Ponds, VIC, Australia
Sheng Gao
Department of Electronic Engineering, Tsinghua University, Beijing, China
Yi Gao
Department of Electronic Engineering, Tsinghua University, Beijing, China
Yiyang Gong
Department of Biomedical Engineering; Department of Neurobiology, Duke University,
Durham, NC, United States

ix
x Contributors

Aldair E. Gongora
Department of Mechanical Engineering, Boston University, Boston, MA, United States
Grace X. Gu
Mechanical Engineering, University of California, Berkeley, Berkeley, CA, United States
Yu Gui-Lan
Beijing Jiaotong University, School of Civil Engineering, Beijing, China
Che-Lun Hung
Institute of Biomedical Informatics, National Yang Ming Chiao Tung University, Taipei, Taiwan
Santiago Muiños Landin
AIMEN Technology Centre, Smart Systems and Smart Manufacturing–Artificial Intelligence and
Data Analytics Laboratory, Pontevedra, Spain
Chihun Lee
Department of Mechanical Engineering, Pohang University of Science and Technology
(POSTECH), Pohang, Republic of Korea
Fushan Li
Institute of Optoelectronic Technology, Fuzhou University, Fuzhou, China
Yu Li
School of Physics, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing;
Silicon Photonics R&D Center, CUMEC, Chongqing, China
Xing Lin
Department of Electronic Engineering, Tsinghua University, Beijing, China
Zhengchang Liu
School of Physics, Academy for Advanced Interdisciplinary Studies, Peking University, Beijing,
China
Cuicui Lu
Beijing Institute of Technology, Beijing, China
Yongfeng Lu
College of Chemistry, Fuzhou University, Fuzhou; Fujian Science & Technology Innovation
Laboratory for Optoelectronic Information, China
G. Milano
Advanced Materials Metrology and Life Sciences Division, Italian National Institute of Metrology
Research, Torino, Italy
Jaebum Noh
Department of Mechanical Engineering, Pohang University of Science and Technology
(POSTECH), Pohang, Republic of Korea
Ravi Pradip
Universitat Politècnica De Catalunya, Barcelona, Spain
Junsuk Rho
Department of Mechanical Engineering; Department of Chemical Engineering, Pohang
University of Science and Technology (POSTECH), Pohang, Republic of Korea
 Contributors xi

C. Ricciardi
Applied Science and Technology Department (DISAT), Politecnico di Torino, Torino, Italy
Verda Saygin
Department of Mechanical Engineering, Boston University, Boston, MA, United States
Kelsey L. Snapp
Department of Mechanical Engineering, Boston University, Boston, MA, United States
Sunae So
Department of Mechanical Engineering; Graduate School of Artificial Intelligence, Pohang
University of Science and Technology (POSTECH), Pohang, Republic of Korea
Zhongwei Xu
Institute of Optoelectronic Technology, Fuzhou University, Fuzhou, China
Charles Yang
Mechanical Engineering, University of California, Berkeley, Berkeley, CA, United States
Lehan Yao
Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign,
Urbana, IL, United States
Hongyi Yuan
Beijing Institute of Technology, Beijing, China
Jooyeong Yun
Department of Mechanical Engineering, Pohang University of Science and Technology
(POSTECH), Pohang, Republic of Korea
Nianen Zhang
Beijing Institute of Technology, Beijing, China
Zhizhou Zhang
Mechanical Engineering, University of California, Berkeley, Berkeley, CA, United States
Yuanhui Zheng
College of Chemistry, Fuzhou University, Fuzhou; Fujian Science & Technology Innovation
Laboratory for Optoelectronic Information, China
Preface

Nanoscience has been an active research field over the past several decades, triggered
by the rising interest in exploring new science and technologies with materials at the
nanoscale. Artificial intelligence (AI) has led to revolutionary development in many
applications such as communication, computer vision, and speech/image recognition.
Bridging the link between nanoscience and AI can provide a new generation of technol-
ogies that will have large impacts on our society. Some of the AI tools are being
implemented by researchers to advance nanoscience and nanotechnology through
AI-enhanced design, manufacturing, measurements, and applications of nanomaterials
and nanodevices. Meanwhile, nanotechnology provides enormous data sets and new
hardware platforms to advance the training of Al models and the implementation of
AI algorithms.
With great contributions from some of the top researchers working in the interdis-
ciplinary field, Intelligent Nanotechnology: Merging Nanoscience and Artificial Intelligence
provides an overview of the advances in science and technology made possible by the
convergence of nanoscience and AI. Section 1 focuses on AI-enhanced design, charac-
terization and manufacturing of nanomaterials, and the use of AI to improve important
material properties, with an emphasis on mechanical, photonic, phononic, electronic,
and magnetic properties. The use of AI in the acquisition and analysis of data in exper-
iments is explored as well. Section 2 includes chapters on AI technologies that have been
enhanced through nanotechnology platforms. For example, memristors, neuromorphic
computing, and artificial neural networks are included. Finally, Section 3 reviews the
advances in applications enabled by the merging of nanotechnology and artificial
intelligence, including examples from biomedicine, chemistry, anticounterfeiting, and
automated research. We hope that this book will provide a bridge to further collabora-
tions among researchers from different fields.
Yuebing Zheng
Zilong Wu

xiii
Acknowledgments

The editors acknowledge the financial support from the National Science Foundation
(NSF-ECCS-2001650) and the National Institute of General Medical Sciences of the
National Institutes of Health (DP2GM128446).

xv
CHAPTER 1

Inverse design meets nanophotonics:

From computational optimization to
artificial neural network
Jaebum Noha,*, Trevon Badloea,*, Chihun Leea,*, Jooyeong Yuna,*, Sunae Soa,b,*,
and
a
Junsuk Rhoa,c
Department of Mechanical Engineering, Pohang University of Science and Technology (POSTECH), Pohang,
Republic of Korea
b
Graduate School of Artificial Intelligence, Pohang University of Science and Technology (POSTECH), Pohang,
Republic of Korea
c
Department of Chemical Engineering, Pohang University of Science and Technology (POSTECH), Pohang, Republic of Korea

1. Computational inverse design

The methods for discovering optimized designs for photonic platforms using computa-
tional algorithms broadly fall into one of two categories, namely, gradient-based or
evolutionary-based algorithms. Indeed, neural networks are trained using gradients
and backpropagation, so are fundamentally related in this way. However, a key difference
occurs in the fact that traditional computation inverse design methods are generally
defined by a specific objective function to be optimized. This could be any required opti-
cal property, for example, a resonance at a specific wavelength or maximizing the broad-
band efficiency of a device. The design space can be as free or as limited as the researcher
requires, which could be directly related to the fabrication constraints of certain materials
or techniques. In this section, we will focus on the application of computational optimi-
zation without the integration of neural networks as an introduction to computational
methods. We will describe the overall goals and methodologies of gradient and
evolutionary-based algorithms and provide examples of recent results that showcase
the power of numerical optimization techniques in the design of nanophotonic devices.
A simple visual description of each method is provided in Fig. 1.

1.1 Gradient-based
Gradient-based optimization techniques rely on the computation of the local gradient of
the objective function in relation to the designated design parameters. This gradient then
guides the optimization algorithm in a certain direction toward a minimum (or maxi-
mum) for the objective function. A problem arises in that there is no guarantee that this
local point of inflection in the gradients of the objective function is the global minimum,

Intelligent Nanotechnology Copyright © 2023 Elsevier Inc.

https://doi.org/10.1016/B978-0-323-85796-3.00001-9 All rights reserved. 3
4 Intelligent nanotechnology

Fig. 1 Visual representations of numerical optimization methods. (A) Gradient descent method. The
blue (gray in the print version) and red (dark gray in the print version) arrows represent optimization
paths from different starting points, leading to different local minima. (B) Genetic algorithm. The
population is made up of N unique solutions to the problem, here with six different parameters.
After each has been given a score and ranked in terms of the fitness function, the best solutions
are used to perform crossover, where the genes of two-parent solutions are spliced together to
create a new child solution. With a specified random chance, some solutions also undergo
mutation to create new solutions. This allows for the optimization to have some chance of
escaping local minima. (C) Particle swarm. Three particles (orange circles (gray in the print version))
with a certain position and momentum explore the parameter space independently and are
updated based on their local information as well as the information of the population as a whole,
to hopefully end in the global minimum, i.e., the optimal solution to the problem. (For
interpretation of the references to color in this figure legend, the reader is referred to the web
version of this article.)

which is, in general, what we wish to uncover to produce the most optimized design. If
the objective function is convex, however, all local minima are also global minima,
meaning that globally optimized solutions can be found. As described in Fig. 1A, after
the gradient has been calculated at a specific location, the optimization algorithm takes
a step in that direction. We need to determine how big of a step to take in the given
direction, and how frequently to measure the gradient. The size of the step is an impor-
tant factor, as too large of a step could lead to overshooting the minimum, while the fre-
quency of measuring the gradient is determined by computational limitations. Of course,
ideally, we would like to know the gradient at every point with infinitesimally small steps
in order to uncover the optimum global solution, but this of obviously unrealistic in prac-
tice. Another important factor in gradient-based optimization methods is the starting
guess in the parameter space. Due to these limitations, it is common to do multiple runs
 Inverse design meets nanophotonics 5

with different starting points and step sizes, increasing the computational load. In an
attempt to lighten the computational load on calculating the gradients of the objective
function of high-dimensional parameter’s spaces, the adjoint method has been intro-
duced. This allows the gradients to be calculated using only a singular additional adjoint
simulation, significantly reducing the computational load [1,2].

1.1.1 Adjoint optimization

Since the goal of optimization is to find the parameters in the design space that maximizes
the achievable performance of the given objective, the traditional method of designing
individual structures and arranging them in periodic arrays is unnecessary. In addition, by
treating the device as a whole, with the only constraint being on the final output objec-
tive, issues such as coupling and interactions between neighboring structures are eradi-
cated. This can allow adjoint optimization to find structures and layouts that avoid
suboptimal designs. There have been various applications of adjoint optimization in
the field of nanophotonics for designing large-scale and multilayered devices for appli-
cations such as waveguides [3–5] and grating couplers [6,7]. Some of the recent demon-
strations will be introduced and the results of the devices designed using adjoint
optimization will be discussed in this section.
Adjoint optimization allows for only one additional simulation in order to calculate
the gradients of the objective function with regard to the design parameters. This allows
for numerous structures to be optimized in a single pass. Using this framework, Backer
numerically designed a double metalens system, holographic element, and an all-optical
neural network, proving the versatility and robustness of adjoint optimization (Fig. 2A
and B) [8]. Using a square TiO2 nanostructure on the glass library, all of the aforemen-
tioned applications were realized. The achromatic, polarization-independent, doublet
metalens was designed with an aperture of 800 μm and a back focal length of 2 mm. It
showed a bandwidth of 160 nm across the visible spectrum. To confine the design param-
eter space to ensure that the nanostructures can be realistically fabricated, limits of the
minimum and maximum size of the nanostructures were enforced. If the optimization
gradient specified an unrealistic structure that is outside of the designated limits, then
it was truncated to the limit. Full-wave 3D simulations with FDTD are computationally
expensive, so all the propagations in both the forward and adjoint simulations were car-
ried out in a 2D cross section of the total metalens to produce a library of results, and
Fourier optics was used to propagate the resulting waves to the far field. Since the concept
of the adjoint method is similar to backpropagation methods used in conventional
machine learning algorithms, the system can also produce optical neural networks. Using
an array of 10 detectors, it was shown that an adjoint optimized metasurface can success-
fully classify images of numbers with an accuracy of 84%. Unfortunately, these results
were purely numerical; however, the power that adjoint optimization can bring to mul-
tiple tasks with large parameter spaces was clearly proven.
6 Intelligent nanotechnology

Fig. 2 Examples of nanophotonic devices designed using the adjoint method. (A) Inverse design of an
achromatic doublet metalens. (i) Schematic of the design. (ii) The widths of the structures that make up
the metalens. (iii) Phase profile across metasurface apertures at λ ¼ 640 nm. (iv) Cross section of focal
plane intensity at all designed wavelengths. (v) The optical efficiency of the metalens. (B) Optical neural
network for handwritten digit classification. (i) Schematic of the design. (ii) Simulated test inputs and
outputs of the optimized design. (iii) Confusion matrix for the MNIST dataset. (iv) The optimized
metasurface designs. (C) A multiwavelength metalens design. (i) Schematic representation of the
forward, (ii) the adjoint simulations in the adjoint optimization technique, and (iii) color maps of
the surface electric current densities used as excitation sources in the adjoint simulations. Panels
A and B are adapted with permission from Mansouree, M., et al., Multifunctional 2.5D metastructures
enabled by adjoint optimization. Optica 7 (2020) 77–84, copyright 2020 Optical Society of America.
Panel C is adapted with permission from Backer, A.S., Computational inverse design for cascaded
systems of metasurface optics. Opt. Express 27 (2019) 30308–30331, copyright 2019 Optical Society of
America.

In another demonstration of the adjoint optimization technique, the design spaces of

metasurfaces have been increased from their general two-dimensional limitations, into
2.5D devices [9]. A 2.5D metalens that focuses different wavelengths of light at different
focal points was designed and fabricated. By including the layers of metasurfaces into the
design, the complexity is increased dramatically, as interactions between neighboring
nanostructures in the same plane are also joined by interactions with the layers above
and below. This could be an extremely arduous design to optimize for a researcher to
optimize by hand; however, despite the increased degrees of freedom and complexity
of the design space, utilization of the adjoint optimization technique produced devices
with considerably higher efficiencies than those designed using conventional methods.
 Inverse design meets nanophotonics 7

An FDTD solver was used to simulate the electric fields of plane waves of the desired
wavelengths incident on the metasurface and the corresponding adjoint simulations,
which are defined as the same structure illuminated by sources that are equal to the com-
plex conjugate of the desired output fields (Fig. 2C). This can be understood as propa-
gating the desired output fields back toward the metasurface. Through the calculated
gradients, the dimensions of nanoposts were manipulated to find the optimized param-
eters. This was proved for both two and three layers of metasurfaces, with focusing effi-
ciencies of over 50%. As the fabrication process of multilayer metasurfaces is much more
involved, the performance increase must be worthwhile in practice. Here, the increase of
focusing efficiency from 35% to 48%, to 60% for the single, bilayer, and trilayer meta-
surfaces proved the added complexity of fabrication is indeed worth the extra effort in
order to produce optimized devices.
Expansions of adjoint optimization include using improved update rules [10,11],
automatic differentiation [12], and extension into the nonlinear regime [13]. Limitations
of adjoint optimization come from the discretization of the structure and the meshing
accuracy of the full-wave simulations. Higher accuracy comes with a higher computa-
tional cost for both the forward and adjoint simulations. Although this optimization
technique does indeed simultaneously optimize the whole device at each update step,
full-wave simulations of large-scale devices require servers or clusters with large enough
memory, or other simulation tricks must be utilized, such as making use of symmetry or
simulating a 2D cross section.

1.1.2 Topology optimization

Topology optimization (TO) has recently been introduced to the field of nanophotonics,
after being widely used in numerous fields for the design of physical systems and mechan-
ical structures [14,15]. It is defined by the allocation of material either existing or not at
specific points in a defined area, in order to satisfy an objective function. This means that
each grid point is a design parameter, leading to the geometry being optimized in a pixel-
like fashion. Again, in order to update the design over each iteration, a gradient needs to
be calculated in relation to the parameter space, so TO is often coupled with the com-
putationally efficient adjoint method, as described in the previous section. TO was intro-
duced to the field of nanophotonics to design low-loss waveguide bends [16] and is now a
staple method for the optimization of a magnitude of different nanophotonic devices
[17]. See Ref. [18] for an in-depth review and description of the fundamentals and pro-
cedures of the optimization method, and Ref. [19] for a recent tutorial that includes FEM
model examples.
Using TO, Ref. [20] designed reconfigurable metalenses that focus light at wavelengths
that are an order of magnitude apart, and also an active metalens that shifts their focal length
depending on the state of an active material. This was achieved using a 10-layer, multi-
height TO method that has around 10 [4] degrees of freedom (Fig. 3A). By modeling
8 Intelligent nanotechnology

the problem in an axisymmetric domain using Maxwell’s equations with a time-harmonic

behavior, the full 3D equations were replaced with their simpler 2D counterparts. This
allows the computational complexity to be reduced at the sacrifice of restricting the
geometric freedom of the device. To further simplify the procedure, far-field transforma-
tions were applied rather than simulating the entire spatial domain between the lens and the
focal point. The figure of merit was defined as the intensity of the electric field at the focal
point, and the gradients were calculated using adjoint sensitivity analysis [21]. As an
experimental validation of the technique, a single-layer metalens made up of variable
heights was fabricated. The final design focused light with a wavelength of 1550 nm with
a numerical aperture of 0.4 with an absolute power transmission of around 93%. Although
this work shows the promise of TO, in reality, the fabrication of a 10-layer metasurface
with variable heights is a great challenge, but one that could become a reality with
improved nanofabrication methods such as 2.5D lithography or multi-photon
polymerization.
Other work has shown that TO can be used to tailor the phase front of the light at
will. In particular, using freeform optimization, Sell et al., proved numerically and exper-
imentally that the optimization of a device for multiple wavelengths at the same time is
possible [22] (Fig. 3B). In principle, the technique can be scaled to a large number of
wavelengths at the cost of added computational effort. Using silicon, a beam steering
device capable of steering N wavelengths into N unique diffraction orders, with an effi-
ciency that scales as 1/N0.5 was designed. The optimization is carried out over the whole
periodicity of the metasurface, meaning that the space that is being optimized is much
larger than the wavelength it is being designed for. This is in stark contrast to the con-
ventional method of designing subwavelength nanoantennas to build the total surface.
This also allows the freedom of the structures to be of any shape that is required, without
any constraint on local periodicities. This freedom in design space allows for curves and
interconnected geometries that produce optical coupling and resonances that would oth-
erwise be nearly impossible to design. One limitation of this work was the use of rigorous
coupled-wave analysis (RCWA) solver for the simulations, as it requires the structures to
be periodic. This could easily be overcome through the use of other methods of solving
Maxwell’s equations that do not rely on periodic boundary conditions [23]. Examples of
multiwavelength and large-angle aberration-free metalenses have also been demonstrated
using TO [24].
TO is able to uncover designs for optical properties that are very rare to find in nature.
In particular, Shi et al. produced single-layered optical elements that show a considerable
birefringence for different polarizations of incident light [25] (Fig. 3C). Using TO allows
the device to have arbitrary geometries, which allow the degrees of freedom to be as big as
the resolution of the individual pixels to be optimized. Elliptical birefringence is achieved
in a single planar device by exploiting the incident angle of the light to break the sym-
metry of the structure.
Fig. 3 Examples of nanophotonic devices designed using topology optimization. (A) Normalized electric field in the x–z plane through the center
of the metalens for (i) ¼ 1 μm and (ii) ¼ 10 μm. (iii) The power in the z-direction through the focal plane normalized to the maximum of the Airy
disc for ¼ 1 μm and (iv) ¼10 μm. (v) A 3D illustration of the 10-layer metalens design. (vi) Cross section of single-layer reference design.
(B) Experimental characterization of the broadband blazed metagrating. (i) Scanning electron microscopy image of the blazed metagrating
operating between 1000 and 1300 nm for TM-polarized normally incident waves. (ii) Plot of the theoretical and experimental efficiencies vs
wavelength for the device. (C) (i) Schematic of the optimized unit structure. (ii) SEM image of a fabricated sample. Scale bar, 1 μm. (iii) The
arrows represent the angle-dependent eigen-polarization states of the device. Different colors correspond to different angles of incidence. As
the angle of incidence varies, the device can be continuously tuned between linear and elliptical birefringence. (iv) Angle-dependent
polarization generation. For a fixed incident polarization, the output polarization state changes continuously from right circular polarization
through horizontal linear polarization to 45 degree linear polarization for varying angles of incidence. Panel A is adapted with permission
from Christiansen, R.E., et al., Fullwave maxwell inverse design of axisymmetric, tunable, and multi-scale multi-wavelength metalenses. Opt.
Express. 28 (2020) 33854–33868, copyright 2020 Optical Society of America. Panel B is adapted with permission from Sell, D., Yang, J., Doshay, S.,
Fan, J.A., Periodic dielectric metasurfaces with high-efficiency, multiwavelength functionalities. Adv. Opt. Mater. 5 (2017) 1700645, copyright 2017
John Wiley and Sons. Panel C is adapted with permission from Shi, Z., et al., Continuous angle-tunable birefringence with freeform metasurfaces
for arbitrary polarization conversion. Sci. Adv. 6 (2020) eaba3367, copyright 2020 CC BY-NC.
10 Intelligent nanotechnology

Contrary to intuition, it has been shown that random initial geometries produce bet-
ter results than using existing nanophotonic designs as the starting point [26]. TO usually
has no constraints on the constituent parts of the final design, which could possibly lead to
unrealistic designs. For example, if there are voids or floating elements, this may produce
an output that maximizes the objective function, but it would be impossible to realize
experimentally. Therefore, consideration of fabrication feasibility has also become a
key point of improving devices designed using TO [27].

1.2 Evolutionary algorithms

In contrast to gradient-based optimization methods, evolutionary algorithms take inspi-
ration from nature to reach designs that meet a certain design objective. The most well-
known and often used examples of evolutionary algorithms are the genetic algorithm
(GA) and particle swarm optimization (PSO). Both are known as metaheuristic
approaches, meaning that they are able to find a reasonable solution to an optimization
problem with limited information or computational cost. Although there is limited
mathematical evidence that evolutionary algorithms converge to global optima, they
do often reach extremely good results [28,29].

1.2.1 Genetic algorithm

The GA is inspired by Charles Darwin’s theory of natural evolution, by focusing on the
selection, mutation, and crossover of a population of solutions. A visual representation is
shown in Fig. 1B. The optimization relies on the knowledge of good solutions within
the population, in order to reach better solutions, while also keeping an amount of ran-
dom mutation to allow for new solutions outside of the current direction to be accessible.
The GA starts with a population of possible solutions, in nanophotonic design, this could
be a set of geometric parameters for each individual. Following the lead of natural selec-
tion, the parameters are called genes, while a set of genes that form the solution for one
individual is known as a chromosome. Each solution is then ranked in terms of its per-
formance, with higher scores having a higher chance of being selected for the next iteration
or generation. Crossover is another important part of the GA, whereby the genes of parent
solutions are mixed together to form new solutions. This allows new child solutions to be
formed from high-performance solutions, which should result in new, better solutions.
Mutation relates to some genes being changed at random, allowing for a measure of diver-
sity in the population, and avoiding premature convergence to a possible suboptimal
solution. The important parameters of the GA are the population size, informative fitness
function, and the amount of mutation and crossover. The population size is usually lim-
ited by the available computational resources, so it needs to be chosen to appropriately
find a good solution in a reasonable amount of time. This simple yet intuitive technique
has been applied to the field of nanophotonics with great success to design waveguides
and photonic crystals [30], and some examples [31] will be discussed here.
 Inverse design meets nanophotonics 11

In Ref. [32] the GA was utilized with some modifications in order to design phase-
controlled metasurfaces. In this work, the concept of the crossover was omitted
completely, in its place only mutation ensued. In order to reach an optimal solution,
the amount of mutation was controlled to be smaller and smaller as the iterations were
concluded. An elaborate fitness function was produced to create the exact desired
response for the final solution. Furthermore, physical constraints were included to ensure
that unphysical results were eradicated during the optimization process. This kind of
supervision of the mutation is essential to keep the results in the desired domain. In
another work, strong chiral dichroism in a plasmonic metasurface was demonstrated,
using a micro-genetic algorithm [33]. The key feature of this adaptation is that instead
of having mutation, there is a complete repopulation of the worst-performing solutions
with completely new random solutions for faster convergence. Since the GA has no
bounds in terms of the number of parameters that it can optimize in a single iteration,
it owes well to freeform designs where the surface is discretized into individual pixels
that can be chosen to include or exclude material, similar to the TO method [34].
But since no gradients are needed, there is no limitation on the fitness function to
be differentiable. Here, the surface was discretized into a binary pattern as shown in
Fig. 4A. Circular dichroism of 0.63 and 0.60 was reported for fabricated samples. The
GA has also been used to choose the photonic elements from a library in order to produce
the desired response [35]. The design of the individual structures was done by hand, but
the phase profile was achieved using the optimization power of the GA. Using an adap-
tive GA allows for the problem to be split into subproblems that are solved in sequence, to
reach the final solution [36]. The authors note that the large number of solutions that are
produced with results that are close to the designed fitness function could also be
employed later in data hungry methods such as a deep learning models. In a final example
of using the GA to optimize the performance of nanophotonic applications, it was used to
design perfect absorbers and color filters for single nanostructures and unit cells with mul-
tiple nanostructures (Fig. 4B) [37]. This study showed the limitations that single nano-
particles have when trying to optimize a solution with limited degrees of freedom. The
colors achieved after increasing the unit cell to include multiple structures were greatly
improved, as were the broadband absorption properties. For all of the examples presented
here, it is interesting for researchers to delve into the underlying physics of the optimized
solution to understand why certain designs are better than others, which in turn, helps
researchers to uncover new ideas that can be applied successfully in new applications.

1.2.2 Particle swarm optimization

Particle swarm optimization gets its influence from the movement of birds and schools of
fish. Each solution is treated as a particle that has a position and a momentum that is
updated at each iteration. The update depends on the particular solution’s local optimum,
as well as the optima of the whole population as a whole, as demonstrated in Fig. 1C [38].
Fig. 4 Examples of nanophotonic devices designed using genetic algorithms. (A) (i) Schematic of the binary pattern for the chiral plasmonic
metasurface. (ii) Top view of the pattern. (iii) SEM image of the fabricated chiral metasurface. (iv) Experimental and simulated absorption
spectra under LCP and RCP at normal incidence. (v) Normalized electric field distributions at the resonant wavelength of 1.62 μm under LCP
and RCP incidence. (B) (i) The distance between the target and designed color generated in the CIE diagram as a function of generations.
Red circles (dark gray in the print version) represent the single-unit cell, green diamonds (gray in the print version) represent the unit cell
with multiple nanostructures. (ii) Comparison of reflection spectra between the optimized samples of single (green (gray in the print
version)) and extended (red (dark gray in the print version)) unit cells. (iii) Comparison of the generated colors and corresponding
configurations. (iv) The generated colors on the CIE diagram. Circles denote the single nanostructure and stars represent the multiple
nanoparticles. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Panel A is adapted with permission from Li, Z., Rosenmann, D., Czaplewski, D.A., Yang, X., Gao, J. Strong circular dichroism in chiral plasmonic
metasurfaces optimized by micro-genetic algorithm. Opt. Express 27 (2019) 28313–28323, copyright 2019 Optical Society of America. Panel B is
adapted with permission from Liu, C., Maier, S.A., Li, G., Genetic-algorithm-aided meta-atom multiplication for improved absorption and
coloration in nanophotonics. ACS Photonics 7 (2020) 1716–1722, copyright 2020 American Chemical Society.
 Inverse design meets nanophotonics 13

The updates are governed by a learning rate that determines how much the velocity of the
particle is changed at each update. This, along with the number of particles provides the
most important parameters that need to be optimized for each problem and considered
with regard to the available computing power.
In one example of using PSO for nanophotonic applications, Ref. [39] produced a
freestanding metasurface of silicon for high-performance beam steering (Fig. 5A).
A particle swarm algorithm was coupled with a full-wave simulation through the FDTD
method. After optimization, both the forward transmissions were greatly increased, and
the power of the deflected beam in the desired direction was improved dramatically.
However, it was noted that the combination of 3D simulations is a huge bottleneck
in this type of technique, as the total time needed for a single optimization was around
850 h. In another work using the PSO method, a silicon nitride Y-junction was devel-
oped with the goal of converting linearly polarized light into its opposite polarization
(Fig. 5B) [40]. Conversion efficiencies of over 97% for both polarizations were reported
and confirmed with numerical simulations. Many other examples have also been dem-
onstrated, such as waveguides [41], and pixelated surfaces [42] with multiple objective
functions [43].

Fig. 5 Examples of nanophotonic devices designed using particle swarm. (A) (i) Schematic of a
freestanding silicon nanohole array metasurface acting as a beam deflector with a deflection angle θ.
(ii) Top-down view of the silicon nanohole array beam deflector metasurface constructed using
repeating supercells. (B) Conversion efficiency as a function of the length L for the designed mode-
conversion section and the adiabatic taper structure. Panel A is adapted with permission from Ong, J.R.,
Chu, H.S., Chen, V.H., Zhu, A.Y., Genevet, P. Freestanding dielectric nanohole array metasurface for
mid-infrared wavelength applications. Opt. Lett. 42 (2017) 2639–2642, copyright 2017 Optical Society of
America. Panel B is adapted with permission from Zhang, B., et al., Particle swarm optimized polarization
beam splitter using metasurface-assisted silicon nitride Y-junction for mid-infrared wavelengths. Opt.
Commun. 451 (2019) 186–191, copyright 2019 Elsevier.
Another random document with
no related content on Scribd:
"What America owed abroad can never be computed; it is enough
that it reached an enormous sum, to refund which, even under
favorable circumstances, would have taken years of effort;
actually forced payment brought the nation to the brink of a
convulsion. Perhaps no people ever faced such an emergency and
paid, without recourse to war. America triumphed through her
inventive and administrative genius. Brought to a white heat
under compression, the industrial system of the Union suddenly
fused into a homogeneous mass. One day, without warning, the
gigantic mechanism operated, and two hemispheres vibrated with
the shock. In March, 1897, the vast consolidation of mines,
foundries, railroads, and steamship companies, centralized at
Pittsburg, began producing steel rails at $18 the ton, and at
a bound America bestrode the world. She had won her great
wager with fate. … The end seems only a question of time.
Europe is doomed not only to buy her raw material abroad, but
to pay the cost of transport. And Europe knew this
instinctively in March, 1897, and nerved herself for
resistance. Her best hope, next to a victorious war, lay in
imitating America, and in organizing a system of
transportation which would open up the East.

"Carnegie achieved the new industrial revolution in March,

1897. Within a twelvemonth the rival nations had emptied
themselves upon the shore of the Yellow Sea. In November
Germany seized Kiao-chau, a month later the Russians occupied
Port Arthur, and the following April the English appropriated
Wei-hai-wei; but the fact to remember is that just 400 miles
inland, due west of Kiao-chau, lies Tszechau, the centre,
according to Richthofen, of the richest coal and iron deposits
in existence. There with the rude methods used by the Chinese,
coal actually sells at 13 cents the ton. Thus it has come to
pass that the problem now being attacked by all the statesmen,
soldiers, scientific men, and engineers of the two eastern
continents is whether Russia, Germany, France, England, and
Japan, combined or separately, can ever bring these resources
on the market in competition with the United States."
 B. Adams,
The New Industrial Revolution
(Atlantic Monthly, February, 1901).

UNITED STATES OF AMERICA: A. D. 1897 (January-May).

Arbitration Treaty with Great Britain rejected by the Senate.

See, (in this volume),

VENEZUELA: A. D. 1896-1899.

The correspondence which took place between the governments of

the United States and Great Britain, on the subject of an
arbitration of the Venezuela Boundary dispute, having led to
the revival of a project for the negotiation of a general
treaty of arbitration, which the late American Secretary of
State, Mr. Gresham, had broached to the British government in
the spring of 1895, the terms of such an arrangement were
carefully and fully discussed between Secretary Olney and Lord
Salisbury, during the year 1896, and an agreement was reached
which took form in a solemn compact for the settlement by
arbitration of all matters in difference between the two
countries, signed at Washington on the 11th of January, 1897.
The treaty thus framed was as follows:

"ARTICLE I.
The High Contracting Parties agree to submit to Arbitration in
accordance with the provisions and subject to the limitations
of this Treaty all questions in difference between them which
they may fail to adjust by diplomatic negotiation.

"ARTICLE II.
All pecuniary claims or groups of pecuniary claims which do
not in the aggregate exceed £100,000 in amount, and which do
not involve the determination of territorial claims, shall be
dealt with and decided by an Arbitral Tribunal constituted as
provided in the next following Article. In this Article and in
 Article IV the words 'groups of pecuniary claims' mean
pecuniary claims by one or more persons arising out of the
same transactions or involving the same issues of law and
fact.'

"ARTICLE III.
Each of the High Contracting Parties shall nominate one
arbitrator who shall be a jurist of repute and the two
arbitrators so nominated shall within two months of the date
of their nomination select an umpire. In case they shall fail
to do so within the limit of time above mentioned, the umpire
shall be appointed by agreement between the Members for the
time being of the Supreme Court of the United States and the
Members for the time being of the Judicial Committee of the
Privy Council in Great Britain, each nominating body acting by
a majority. In case they shall fail to agree upon an umpire
within three months of the date of an application made to them
in that behalf by the High Contracting Parties or either of
them, the umpire shall be selected in the manner provided for
in Article X. The person so selected shall be the President of
the Tribunal and the award of the majority of the Members
thereof shall be final.

{578}

"ARTICLE IV.
All pecuniary claims or groups of pecuniary claims which shall
exceed £100,000 in amount and all other matters in difference,
in respect of which either of the High Contracting Parties
shall have rights against the other under Treaty or otherwise,
provided that such matters in difference do not involve the
determination of territorial claims, shall be dealt with and
decided by an Arbitral Tribunal, constituted as provided in
the next following Article.

"ARTICLE V.
Any subject of Arbitration described in Article IV shall be
submitted to the Tribunal provided for by Article III, the
award of which Tribunal, if unanimous, shall be final. If not
unanimous either of the High Contracting Parties may within
six months from the date of the award demand a review thereof.
In such case the matter in controversy shall be submitted to
an Arbitral Tribunal consisting of five jurists of repute, no
one of whom shall have been a member of the Tribunal whose
award is to be reviewed and who shall be selected as follows,
viz:—two by each of the High Contracting Parties, and one, to
act as umpire, by the four thus nominated and to be chosen
within three months after the date of their nomination. In
case they shall fail to choose an umpire within the limit of
time above-mentioned, the umpire shall be appointed by
agreement between the Nominating Bodies designated in Article
III acting in the manner therein provided. In case they shall
fail to agree upon an umpire within three months of the date
of an application made to them in that behalf by the High
Contracting Parties or either of them, the umpire shall be
selected in the manner provided for in Article X. The person
so selected shall be the President of the Tribunal and the
award of the majority of the Members thereof shall be final.

"ARTICLE VI.
Any controversy which shall involve the determination of
territorial claims shall be submitted to a Tribunal composed
of six members, three of whom (subject to the provisions of
Article VIII) shall be Judges of the Supreme Court of the
United States or Justices of the Circuit Courts to be
nominated by the President of the United States, and the other
three of whom (subject to the provisions of Article VIII)
shall be Judges of the British Supreme Court of Judicature or
Members of the Judicial Committee of the Privy Council to be
nominated by Her Britannic Majesty, whose award by a majority
of not less than five to one shall be final. In case of an
award made by less than the prescribed majority, the award
shall also be final unless either Power shall, within three
months after the award has been reported, protest that the
same is erroneous, in which case the award shall be of no
validity. In the event of an award made by less than the
prescribed majority and protested as above provided, or if the
members of the Arbitral Tribunal shall be equally divided,
there shall be no recourse to hostile measures of any
description until the mediation of one or more friendly Powers
has been invited by one or both of the High Contracting
Parties.

"ARTICLE VII.
Objections to the jurisdiction of an Arbitral Tribunal
constituted under this Treaty shall not be taken except as
provided in this Article. If before the close of the hearing
upon a claim submitted to an Arbitral Tribunal constituted
under Article III or Article V either of the High Contracting
Parties shall move such Tribunal to decide, and thereupon it
shall decide that the determination of such claim necessarily
involves the decision of a disputed question of principle of
grave general importance affecting the national rights of such
party as distinguished from the private rights whereof it is
merely the international representative, the jurisdiction of
such Arbitral Tribunal over such claim shall cease and the
same shall be dealt with by arbitration under Article VI.

"ARTICLE VIII.
In cases where the question involved is one which concerns a
particular State or Territory of the United States, it shall
be open to the President of the United States to appoint a
judicial officer of such State or Territory to be one of the
Arbitrators under Article III or Article V or Article VI. In
like manner in cases where the question involved is one which
concerns a British Colony or possession, it shall be open to
Her Britannic Majesty to appoint a judicial officer of such
Colony or possession to be one of the Arbitrators under
Article III or Article V or Article VI.

"ARTICLE IX.
Territorial claims in this Treaty shall include all claims to
territory and all claims involving questions of servitudes,
rights of navigation and of access, fisheries and all rights
and interests necessary to the control and enjoyment of the
territory claimed by either of the High Contracting Parties.

"ARTICLE X.
If in any case the nominating bodies designated in Articles
III and V shall fail to agree upon an Umpire in accordance
with the provisions of the said Articles, the Umpire shall be
appointed by His Majesty the King of Sweden and Norway. Either
of the High Contracting Parties, however, may at any time give
notice to the other that, by reason of material changes in
conditions as existing at the date of this Treaty, it is of
opinion that a substitute for His Majesty should be chosen
either for all cases to arise under the Treaty or for a
particular specified case already arisen, and thereupon the
High Contracting Parties shall at once proceed to agree upon
such substitute to act either in all cases to arise under the
Treaty or in the particular case specified as may be indicated
by said notice; provided, however, that such notice shall have
no effect upon an Arbitration already begun by the constitution
of an Arbitral Tribunal under Article III. The High
Contracting Parties shall also at once proceed to nominate a
substitute for His Majesty in the event that His Majesty shall
at any time notify them of his desire to be relieved from the
functions graciously accepted by him under this Treaty either
for all cases to arise thereunder or for any particular
specified case already arisen.

"ARTICLE XI.
In case of the death, absence or incapacity to serve of any
Arbitrator or Umpire, or in the event of any Arbitrator or
Umpire omitting or declining or ceasing to act as such,
another Arbitrator or Umpire shall be forthwith appointed in
his place and stead in the manner provided for with regard to
the original appointment.
{579}

"ARTICLE XII.
Each Government shall pay its own agent and provide for the
proper remuneration of the counsel employed by it and of the
Arbitrators appointed by it and for the expense of preparing
and submitting its case to the Arbitral Tribunal. All other
expenses connected with any Arbitration shall be defrayed by
the two Governments in equal moieties. Provided, however,
that, if in any case the essential matter of difference
submitted to arbitration is the right of one of the High
Contracting Parties to receive disavowals of or apologies for
acts or defaults of the other not resulting in substantial
pecuniary injury, the Arbitral Tribunal finally disposing of
the said matter shall direct whether any of the expenses of
the successful party shall be borne by the unsuccessful party,
and if so to what extent.

"ARTICLE XIII.
The time and place of meeting of an Arbitral Tribunal and all
arrangements for the hearing and all questions of procedure
shall be decided by the Tribunal itself. Each Arbitral
Tribunal shall keep a correct record of its proceedings and
may appoint and employ all necessary officers and agents. The
decision of the Tribunal shall, if possible, be made within
three months from the close of the arguments on both sides. It
shall be made in writing and dated and shall be signed by the
Arbitrators who may assent to it. The decision shall be in
duplicate, one copy whereof shall be delivered to each of the
High Contracting Parties through their respective agents.

"ARTICLE XIV.
This Treaty shall remain in force for five years from the date
at which it shall come into operation, and further until the
expiration of twelve months after either of the High
Contracting Parties shall have given notice to the other of
its wish to terminate the same.

"ARTICLE XV.
The present Treaty shall be duly ratified by the President of
the United States of America, by and with the advice and
consent of the Senate thereof, and by Her Britannic Majesty;
and the mutual exchange of ratifications shall take place in
Washington or in London within six months of the date hereof
or earlier if possible."

United States, 54th Congress, 2d Session,

Senate Document Number 63.

Public feeling in both countries gave joyful welcome to this

nobly conceived treaty when it was announced. All that was
best in English sentiment and American sentiment had been
shuddering at the thought of possible war between the kindred
peoples, and thanked God for what promised some certitude that
no dispute would be pushed to that barbarous appeal. Only the
mean thought and temper of either country was provoked to
opposition; but, unhappily, the meaner temper and the narrower
and more ignorant opinion on one side of the sea had been
getting so strong a representation in the United States Senate
as to prove capable of much mischief there, on this and other
matters of most serious public concern. When the great
covenant of peace went to that body for approval, there were
senators who found it offensive to them because it came from
the hands of President Cleveland and Secretary Olney; and
there were other senators whose dignity was hurt by the eager
impatience with which the public voice cried out for their
ratifying vote; and still others there were who looked with
official jealousy at the project of an arbitral tribunal which
might sometimes take something from senatorial functions in
foreign affairs. And the combination of pitiful motives had
strength enough to baffle the high hopes and defeat the will
of the American people.
Of the public feeling thus outraged, the following is one
expression of the time, among many which it would be possible
to quote:

Many people "are represented by influential papers like the

St. Paul 'Pioneer Press' and the Minneapolis 'Journal,' the
latter declaring that it is humiliating to think that, widely
as the treaty is favored throughout the country, a few
ill-natured men in the Senate have the power to delay
ratification. In the Central West the feeling is generally
strong for arbitration, if we may judge from the Chicago
'Times-Herald,' the St. Louis 'Republic,' the Indianapolis
'Journal,' and the Cleveland 'Leader.' In the South there are
such cheering reports as this from the Memphis 'Scimetar'; 'If
the treaty now under consideration in the Senate Committee on
Foreign Relations should fail of ratification, public opinion
in this country would demand that the incoming Administration
provide another embodying the same vital principle.' In the
East the sentiment in favor of immediate ratification of the
original draft has been almost universal, the only two
journals of note differing from this being the New York 'Sun'
and the Washington 'Post.' The trend of opinion is shown in
the adoption by the Massachusetts House of Representatives of
an endorsement by a vote of 141 to 11. An important meeting
took place last week in Washington in favor of ratification.
The speakers were ex-Secretary of State Foster, Mr. G. G.
Hubbard, Professor B. L. Whitman, ex-Senator J. B. Henderson,
ex-Governor Stanard, and Justice Brewer, of the Supreme Court.
The last named said; 'I do not believe in saying to the
gentlemen charged with the duty of considering carefully that
treaty, that "you must vote for it." There is something in my
own nature which, when anybody says to me "you must," causes
something to run up my spinal column which says "I won't." It
is the Senate's duty to consider that treaty carefully, and
when I say that, I say it is no trespass upon their rights for
American citizens to express their views of that treaty. What are
the errors and losses incidental to arbitration compared to
the horrors of war? What are a few million dollars of wrongful
damages in comparison to the sacrifice of thousands of human
lives?'"
The Outlook, February 6, 1897.

"This treaty was greeted with widespread favor in the press,

but was antagonized at once in the Senate by the jingo element
and by the personal adversaries of the administration. The
committee on foreign relations reported the draft favorably,
but with certain amendments, on February 1. The ensuing debate
soon revealed that a vote on ratification could not be
obtained before March 4, and the whole matter was dropped. At
the opening of the new Congress the Senate Committee again
considered the treaty and reported it, with amendments, on
March 18. During two weeks' discussion the Senate adopted the
committee's amendments and also others, with the result that
the draft was radically transformed.
{580}
Instead of the general reference of all disputes to the
tribunals, it was provided that any difference 'which, in the
judgment of either power, materially affects its honor or its
domestic or foreign policy,' should be submitted to
arbitration only by special agreement; that no question should
be submitted save with the consent of the Senate in its
treaty-making capacity; and that no claim of a British subject
against a state or territory of the United States should be
submitted under any circumstances. The first of these changes
was due mainly to the objection that without it the Monroe
Doctrine might be subjected to arbitration; the second to the
sensitiveness of senators as to their constitutional functions
in foreign relations; and the third to a desire to protect
states against claims on their defaulted bonds. Other changes
modified materially the method of appointing the arbitrators
for the United States, and struck out entirely the designation
of the King of Sweden as umpire. Even with these amendments,
the opposition to the treaty was not overcome; and the final
vote on ratification, taken May 5, resulted in its rejection,
the vote standing 43 to 26, less than two-thirds in the
affirmative. Thirty Republicans and 13 Democrats voted for the
 treaty; 8 Republicans, 12 Democrats and 6 Populists against
it."

Political Science Quarterly,

June, 1897.

UNITED STATES OF AMERICA: A. D. 1897 (March).

Inauguration of President McKinley.
Leading topics of the inaugural address.
The President's Cabinet.

The inauguration of President McKinley was performed with the

customary ceremonies on the 4th of March. In his inaugural
address, the new President laid somewhat less emphasis than
might have been expected on the need of measures for reforming
the monetary system of the country, but strongly urged that
instant steps be taken to increase the revenues of the
government by a return to higher tariff charges. "With
adequate revenue secured," he argued, "but not until then, we
can enter upon such changes in our fiscal laws as will, while
insuring safety and volume to our money, no longer impose upon
the government the necessity of maintaining so large a gold
reserve, with its attendant and inevitable temptations to
speculation. Most of our financial laws are the outgrowth of
experience and trial, and should not be amended without
investigation and demonstration of the wisdom of the proposed
changes. We must be both 'sure we are right' and 'make haste
slowly.' …

"The question of international bimetallism will have early and

earnest attention. It will be my constant endeavor to secure
it by cooperation with the other great commercial powers of
the world. Until that condition is realized, when the parity
between our gold and silver money springs from and is
supported by the relative value of the two metals, the value
of the silver already coined, and of that which may hereafter
be coined, must be kept constantly at par with gold by every
resource at our command. The credit of the government, the
integrity of its currency, and the inviolability of its
obligations must be preserved. This was the commanding verdict
of the people, and it will not be unheeded.

"Economy is demanded in every branch of the government at all

times, but especially in periods like the present of
depression in business and distress among the people. The
severest economy must be observed in all public expenditures,
and extravagance stopped wherever it is found, and prevented
wherever in the future it may be developed. If the revenues
are to remain as now, the only relief that can come must be
from decreased expenditures. But the present must not become
the permanent condition of the government. It has been our
uniform practice to retire, not increase, our outstanding
obligations; and this policy must again be resumed and
vigorously enforced. Our revenues should always be large
enough to meet with ease and promptness not only our current
needs and the principal and interest of the public debt, but
to make proper and liberal provision for that most deserving
body of public creditors, the soldiers and sailors and the
widows and orphans who are the pensioners of the United
States. …

"A deficiency is inevitable so long as the expenditures of the

government exceed its receipts. It can only be met by loans or
an increased revenue. While a large annual surplus of revenue
may invite waste and extravagance, inadequate revenue creates
distrust and undermines public and private credit. Neither
should be encouraged. Between more loans and more revenue
there ought to be but one opinion. We should have more
revenue, and that without delay, hindrance, or postponement. A
surplus in the treasury created by loans is not a permanent or
safe reliance. It will suffice while it lasts, but it cannot
last long while the outlays of the government are greater than
its receipts, as has been the case during the last two years.
… The best way for the government to maintain its credit is to
 pay as it goes—not by resorting to loans, but by keeping out
of debt—through an adequate income secured by a system of
taxation, external, or internal, or both. It is the settled
policy of the government, pursued from the beginning and
practiced by all parties and administrations, to raise the
bulk of our revenue from taxes upon foreign productions
entering the United States for sale and consumption, and
avoiding, for the most part, every form of direct taxation
except in time of war.

"The country is clearly opposed to any needless additions to

the subjects of internal taxation, and is committed by its
latest popular utterance to the system of tariff taxation.
There can be no misunderstanding either about the principle
upon which this tariff taxation shall be levied. Nothing has
ever been made plainer at a general election than that the
controlling principle in the raising of revenue from duties on
imports is zealous care for American interests and American
labor. The people have declared that such legislation should
be had as will give ample protection and encouragement to the
industries and the development of our country. … The paramount
duty of congress is to stop deficiencies by the restoration of
that protective legislation which has always been the firmest
prop of the treasury. The passage of such a law or laws would
strengthen the credit of the government both at home and
abroad, and go far toward stopping the drain upon the gold
reserve held for the redemption of our currency, which has
been heavy and well-nigh constant for several years. In the
revision of the tariff, especial attention should be given to
the re-enactment and extension of the reciprocity principle of
the law of 1890, under which so great a stimulus was given to
our foreign trade in new and advantageous markets for our
surplus agricultural and manufactured products."

{581}

Without effect, the incoming President urged the ratification

 of the treaty of arbitration with Great Britain, negotiated by
his predecessor and still pending in the Senate. In concluding
his address he announced his intention to convene Congress in
extra session, saying: "The condition of the public treasury
demands the immediate consideration of congress. It alone has
the power to provide revenue for the government. Not to
convene it under such circumstances, I can view in no other
sense than the neglect of a plain duty."

On the day following his inauguration, the President sent to

the Senate the following nominations for his Cabinet, which
were confirmed:

Secretary of State, John Sherman of Ohio;

Secretary of the Treasury, Lyman J. Gage of Illinois;
Secretary of War, Russel A. Alger of Michigan;
Attorney-General, Joseph McKenna of California;
Postmaster-General, James A. Gary of Maryland;
Secretary of the Navy, John D. Long of Massachusetts;
Secretary of the Interior, Cornelius N. Bliss of New York;
Secretary of Agriculture, James Wilson of Iowa.

UNITED STATES OF AMERICA: A. D. 1897 (March-July).

Passage of the Dingley Tariff Act.

Carrying out an intention announced in his Inaugural Address,

President McKinley called Congress together in extra session
on the 15th of March, asking for immediate action to increase
the revenue of the government by increased duties, "so levied
upon foreign products as to preserve the home market, so far
as possible, to our own producers." In his Inaugural Address
the President had expressed the understanding of his party as
to the chief meaning of the late election, by saying that "the
country is … committed by its latest popular utterance to the
system of tariff taxation. … The people have declared that
such legislation should be had as will give ample protection
and encouragement to the industries and development of our
country. … The paramount duty of Congress is to stop
deficiencies by the restoration of that protective legislation
which has always been the firmest prop of the treasury." To
the majority in both Houses of Congress these views were
entirely acceptable, and they were acted upon at once. The
Ways and Means Committee of the House of Representatives in
the previous Congress had already prepared a comprehensive new
tariff bill, which it passed on to its successor. This
ready-made bill was reported to the House on the first day of
the session, by Mr. Dingley, chairman of the newly appointed
committee, as he had been of the one before it. Debate on the
measure began a week later, and was controlled by a fixed
programme, which required it to be ended on the 31st of March.
The bill was then passed, by a vote of 205 against 121. Of the
action of the Senate upon it, and of the main features of the
bill as it was finally shaped and became law, the following is
a succinct account:

"The bill, referred at once to the Senate Committee on

Finance, was reported after a month, on May 8, with important
amendments. There was an attempt to impose some purely revenue
duties, and, as to the protective duties, the tendency was
towards lower rates than in the House bill, though on certain
articles, such as wools of low grade, hides, and others (of
which more will be said presently), the drift was the other
way. The Senate, however, paid much less respect than the
House to the recommendations of the committee in charge. In
the course of two months, from May 4 to July 7, it went over
the tariff bill item by item, amending without restraint,
often in a perfunctory manner, and not infrequently with the
outcome settled by the accident of attendance on the
particular day; on the whole, with a tendency to retain the
higher rates of the House bill. As passed finally by the
Senate on July 7, the bill, though it contained some 872
amendments, followed the plan of the House Committee rather
than that of the Senate Committee. As usual, it went to a
Conference Committee. In the various compromises and
adjustments in the Senate and in the Conference Committee
there was little sign of the deliberate plan and method which
the House had shown, and the details of the act were settled
in no less haphazard fashion than has been the case with other
tariff measures. As patched up by the Conference Committee,
the bill was promptly passed by both branches of Congress, and
became law on July 24. In what manner these political
conditions affected the character of the act will appear from
a consideration of the more important specific changes.

"First and foremost was the reimposition of the duties on

wool. As the repeal of these duties had been the one important
change made by the act of 1894, so their restoration was the
salient feature in the act of 1897. … Clothing wool was
subjected once more to a duty of 11 cents a pound, combing
wool to one of 12 cents. On carpet wool there were new graded
duties, heavier than any ever before levied. If its value was
12 cents a pound or less the duty was 4 cents; if over 12
cents, the duty was 7 cents. … The duties on carpet wool, as
has already been noted, were made higher than ever before. In
the House the rates of the act of 1890 had been retained; but
in the Senate new and higher rates were inserted. … They were
demanded by the Senators from some States in the Far West,
especially from Idaho and Montana. … They [the Senators in
question] needed to be placated and they succeeded in getting
higher duties on the cheap carpet wools, on the plea of
encouragement for the comparatively coarse clothing wool of
their ranches. … The same complications that led to the high
duty on carpet wool brought about a duty on hides. This rawest
of raw materials had been on the free list for just a quarter
of a century, since 1872, when the duty of the war days had
been repealed. … But here, again, the Senators from the
ranching States were able to dictate terms. … In the Senate a
duty of 20 per cent. was tacked on. The rate was reduced to 15
per cent. in the Conference Committee, and so remains in the
act. The restored duties on wool necessarily brought in their
train the old system of high compensating duties on woollens.
 … In the main, the result was a restoration of the rates of
the act of 1890. There was some upward movement almost all
along the line; and the ad valorem duty alone, on the classes
of fabrics which are most largely imported, crept up to 55 per
cent. …

{582}

"On cotton goods the general tendency was to impose duties

lower than those of 1890. This was indicated by the drag-net
rate, on manufactures of cotton not otherwise provided for,
which had been 50 per cent. in 1890, and was 45 per cent. in
1897. On two large classes of textile goods new and distinctly
higher duties were imposed,—on silks and linens. … The mode of
gradation was to levy the duties according to the amount of
pure silk contained in the goods. The duties were fixed by the
pound, being lowest all goods containing a small proportion of
pure silk, and rising as that proportion became larger; with
the proviso that in no case should the duty be less than 50
per cent. … Thus, the duty on certain kinds of silks was $1.30
cents per pound, if they contained 45 per cent in weight of
silk; but advanced suddenly to $2.25, if they contained more
than 45 per cent. … On linens another step of the same kind
was taken, specific duties being substituted here also for
ad-valorem. … Linens were graded somewhat as cottons had been
graded since 1861, according to the fineness of the goods as
indicated by the number of threads to the square inch. If the
number of threads was 60 or less per square inch, the duty was
1¾ cents a square yard; if the threads were between 60 and
120, the duty was 2¾ cents; and so on,—plus 30 per cent.
ad-valorem duty in all cases. But finer linen goods, unless
otherwise specially provided for, were treated leniently. If
the weight was small (less than 4½ ounces per yard), the duty
was but 35 per cent. On the other hand, linen laces, or
articles trimmed with lace or embroidery, were dutiable at 60
percent.,—an advance at 10 per cent. over the rate of 1890. …
It was inevitable, under the political conditions of the
session, that in this schedule something should again be
attempted for the farmer; and, accordingly, we find a
substantial duty on flax. The rate of the act of 1890 was
restored,—3 cents a pound on prepared flax, in place of the
rate of 1½ cents imposed by the act of 1894. …

"On chinaware the rates of 1890 were restored. The duty on the
finer qualities which are chiefly imported had been lowered to
35 per cent. in 1894, and was now once more put at 60 per
cent. On glassware, also, the general ad-valorem rate, which
had been reduced to 35 per cent. in 1894, was again fixed at
45 per cent., as in 1890. Similarly the specific duties on the
cheaper grades of window-glass and plate-glass, which had been
lowered in 1894, were raised to the figures of 1890. … The metal
schedules in the act of 1897 showed in the main a striking
contrast with the textile schedules. Important advances of
duty were made on many textiles, and in some cases rates went
considerably higher even than those of 1890. But on most
metals, and especially on iron and steel, duties were left
very much as they had been in 1894. … On steel rails there was
even a slight reduction from the rate of 1894—$6.72 per ton
instead of $7.84. On coal there was a compromise rate. The
duty had been 75 cents a ton in 1890, and 40 cents in 1894; it
was now fixed at 67 cents. On the other hand, as to certain
manufactures of iron and steel farther advanced beyond the
crude stage, there was a return to rates very similar to those
of 1890. Thus, on pocket cutlery, razors, guns, we find once
more the system of combined ad-valorem and specific duties,
graded according to the value of the article. … Copper
remained on the free list, where it had been put in 1894. …
For good or ill the copper duty had worked out all its effects
years before. On the other hand, the duties on lead and on
lead ore went up to the point at which they stood in 1890.
Here we have once more the signs of concession to the silver
Republicans of the far West. … The duty on tin plate, a bone
of contention under the act of 1890, was disposed of, with
little debate, by the imposition of a comparatively moderate
duty. …

"A part of the act which aroused much public attention and
which had an important bearing on its financial yield was the
sugar schedule—the duties on sugar, raw and refined. … The act
of 1890 had admitted raw sugar free, while that of 1894 had
imposed a duty of 40 per cent. ad valorem. … The price of raw
sugar had maintained its downward tendency; and the duty of 40
per cent. had been equivalent in 1896 to less than one cent a
pound. In the act of 1897 the duty was made specific, and was
practically doubled. Beginning with a rate of one cent a pound
on sugar tested to contain 75 per cent., it advanced by stages
until on sugar testing 95 per cent. (the usual content of
commercial raw sugar) it reached 1.65 cents per pound. The
higher rate thus imposed was certain to yield a considerable
increase of revenue. Much was said also of the protection now
afforded to the beet sugar industry of the West. That
industry, however, was still of small dimensions and uncertain
future. … On refined sugar, the duty was made 1.95 cents per
pound, which, as compared with raw sugar testing 100 per
cent., left a protection for the domestic refiner,—i. e., for
the Sugar 'Trust,'—of 1/8 of one cent a pound. Some intricate
calculation would be necessary to make out whether this
'differential' for the refining interest was more or less than
in the act of 1894; but, having regard to the effect of the
substitution of specific for ad-valorem duties, the Trust was
no more favored by the act of 1897 than by its predecessor,
and even somewhat less favored. The changes which this part of
the tariff act underwent in the two Houses are not without
significance." In the bill passed by the House. "the so-called
differential, or protection to the refiners, was one-eighth of
a cent per pound. In the Senate there was an attempt at
serious amendment. The influence of the Sugar Trust in the
Senate had long been great. How secured, whether through party
contributions, entangling alliances, or coarse bribery, the
public could not know; but certainly great, as the course of
legislation in that body demonstrated." The Senate attempted