Calculating many SMARTIES T-matrices

Below is an example script to run many SMARTIES simulations and export the T-matrices in HDF5. The script is available as standalone script smarties_batch.R.

library(glue)
library(dplyr)
library(purrr)

# read in the template
template <- glue_collapse(readLines('smarties/_template.m'), sep = "\n")

parameters <- expand.grid(a = seq(10, 100, by=5), c = seq(10, 100, by=5),
                          material = c("Au", "Ag", "Si"), 
                          medium = c("vacuum", "water")) |> 
  filter(a != c) |> # use Mie for this
  mutate(shape = ifelse(a > c, "oblate", "prolate"),
         n = ifelse(medium == "water", 1.33, 1.0),
         source = case_when(material == "Au" ~ "Raschke et al 10.1103/PhysRevB.86.235147", 
                            material == "Ag" ~ "Raschke et al 10.1103/PhysRevB.91.235137",
                            material == "Si" ~ "Aspnes et al 10.1103/PhysRevB.27.985",
                            .default = "Unknown material!"))

parameters$step <- 1
nrow(parameters)

[1] 2052

For initial testing, we’ll run much fewer combinations

parameters <- expand.grid(a = seq(20, 50, by=10), c = seq(20, 50, by=10),
                          material = c("Au", "Ag"), 
                          medium = c("water")) |> 
  filter(a != c) |> # use Mie for this
  mutate(shape = ifelse(a > c, "oblate", "prolate"),
         n = ifelse(medium == "water", 1.33, 1.0),
         source = case_when(material == "Au" ~ "Raschke et al 10.1103/PhysRevB.86.235147", 
                            material == "Ag" ~ "Raschke et al 10.1103/PhysRevB.91.235137",
                            material == "Si" ~ "Aspnes et al 10.1103/PhysRevB.27.985",
                            .default = "Unknown material!"))
parameters$step <- 5

Even with these restricted options, there’s already 24 combinations.

a	c	shape	material	medium
30	20	oblate	Au	water
40	20	oblate	Au	water
50	20	oblate	Au	water
20	30	prolate	Au	water
40	30	oblate	Au	water
50	30	oblate	Au	water
20	40	prolate	Au	water
30	40	prolate	Au	water
50	40	oblate	Au	water
20	50	prolate	Au	water
30	50	prolate	Au	water
40	50	prolate	Au	water
30	20	oblate	Ag	water
40	20	oblate	Ag	water
50	20	oblate	Ag	water
20	30	prolate	Ag	water
40	30	oblate	Ag	water
50	30	oblate	Ag	water
20	40	prolate	Ag	water
30	40	prolate	Ag	water
50	40	oblate	Ag	water
20	50	prolate	Ag	water
30	50	prolate	Ag	water
40	50	prolate	Ag	water

We use the “glue” package to inject the parameters into the template, where the variables are indicated between braces {}. The process loops over each parameter combination and outputs a new file with corresponding filename.

write_script <- function(a, c, material, medium, shape, n, step, source){
 script <- glue(template)   
 cat(script, file = glue('smarties/run_{material}_{medium}_{a}_{c}.m'))
 cat(glue("run_{material}_{medium}_{a}_{c}\n\n"), 
     file = "smarties/batch.m", append = TRUE)
}
 cat("%% Running all the files below\n",file = "smarties/batch.m", append = FALSE)
pwalk(rowwise(parameters), write_script)

Running the batch script in Matlab results in 24 output T-matrix files of size between 4 and 6Mb each, with the 5nm step defined above. The value of Lmax needed for \(10^{-10}\) accuracy in \(\langle\sigma_\text{ext}\rangle\) appears to be around 11.