Adaptive Quantum Design for Nanoscience

1
Adaptive Quantum Design for Nanoscience

• Jason Thalken, Stephan Haas, Anthony Levi
• University of Southern California
• Department of Physics and Astronomy

2
Nano-Scale Design

• Quantum effects can not be ignored
• Complex interactions require computationally
  expensive quantum models
• Classical devices will not maintain functionality
  when scaled into this regime
• New functionalities may exist which have no
  counterpart at larger length scales
• Broken-symmetry configurations must be examined
• Breaking symmetry often has effects for which we
  have no a priori intuition
• The desired functionality may result from only a
  small fraction of the nearly infinite set of all
  possible configurations

3
Adaptive Quantum Design

• A useful device functionality is specified by
  humans.
• Computers evaluate the functionality of potential
  designs using an efficient and accurate quantum
  model.
• Advanced search algorithms find optimal design
  solutions to best fit the specified functionality.

It is also possible to remove human input
entirely, allowing machines to search for
solutions which exhibit any useful or
interesting functionality.
4
First Example Density of States of 4 Atoms in 1D
Target DOS 4 equidistant peaks
5
Adaptive Quantum Design 9 Atoms in 2D
(3 3) 2D periodic array density of states
120

• Start with 2D periodic array of atoms.
• Use tight-binding description of electrons
  around atoms.
• Break symmetry of 2D atom array to emulate flat
  density of states.
• Local update guided random walk.

Target density of states is quasi-2D
80
N(E)
40
0
0
-5
5
Energy, E/t
Atom position, y
N(E)
Atom position, x
Energy, E/t
6
Second Example Excitonic Absorption in AlGaAs
Quantum Well Structures
F 0 kV/cm
F 70 kV/cm
Apply an Electric Field
Eg 1.43 eV
Position, z (nm)
Position, z (nm)
Effective Masses Electron 0.067 me, Heavy
hole 0.340 me
7
Effects of Applied Electric Field on Absorption

• When an electric field is applied to a symmetric
  square well, both the absorption peak strength
  and absorbed photon energy diminishes (quantum
  confined Stark effect)

8
Target An Absorption Frequency Switch

• Specifications
• Match absorption strength at 0 and 70 kV/cm
• Separate the two peaks by more than two line
  widths
• Both peaks should have large absorption strength

• A target function represents the desired quantum
  physical model output. In this case, the target
  function is represented by two points of equal
  absorption strength separated in energy by at
  least 0.012 eV
• A fitness function represents the weighted
  distance between the physical models output for
  a particular solution and the target function.
  The most desirable solution will have the lowest
  possible fitness value.

9
Solution Field Induced Ionization

• This solution was discovered using a
  machine-based genetic algorithm search
• Exponential loss in peak strength intensity as
  hole ionizes suggests an intensity modulator can
  be developed from a similar structure

10
A New Approach to DesignAutomated Device
Synthesis

• Motivation
• Removing human input from the design process will
  lift many time and target related limitations
• It is unreasonable to expect humans to perform an
  exhaustive search of n-dimensional configuration
  space

Interesting Solutions
Solutions

Computer Sorting
11
Computer-Sorted Interesting Absorption Paths
12
Conclusions

• Adaptive Quantum Design search for optimum
  system configurations which closely match target
  functions, which leads to the discovery of new
  molecular building blocks.
• New paradigm for nanoscience target dictates
  system shape.
• Removing the target machines that search for
  optimal configurations can perform exploratory
  searches for interesting solutions as well