Merge pull request invrs-io#101 from invrs-io/designs

Include additional metalens designs in docs

Author: mfschubert

docs/notebooks/metalens_challenge.ipynb

@@ -184,15 +184,97 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Metalens optimization\n",
+    "## Performance versus length scale\n",
     "\n",
-    "To optimize a metalens, you may follow the recipe in the `ceviche_challenge` and `metagrating_challenge` notebooks."
+    "Next, we will compute the performance of the reference devices and measure their length scale using the [imageruler](https://github.com/nanocomp/imageruler) package."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import glob\n",
+    "import imageruler\n",
+    "\n",
+    "results = {\"invrsio\": [], \"rasmus\": [], \"mo\": []}\n",
+    "\n",
+    "for fname in glob.glob(\"../../reference_designs/metalens/Ex/*.csv\"):\n",
+    "    design = onp.genfromtxt(fname, delimiter=\",\")\n",
+    "    design = design[:, ::-1]\n",
+    "\n",
+    "    grid_spacing = 0.01 if \"Rasmus\" in fname else 0.02\n",
+    "    spec = dataclasses.replace(\n",
+    "        metalens_challenge.METALENS_SPEC,\n",
+    "        width_lens=design.shape[0] * grid_spacing,\n",
+    "        width_pml=0.5,\n",
+    "        thickness_lens=design.shape[1] * grid_spacing,\n",
+    "        grid_spacing=grid_spacing,\n",
+    "    )\n",
+    "    challenge = challenges.metalens(spec=spec)\n",
+    "    dummy_params = challenge.component.init(jax.random.PRNGKey(0))\n",
+    "\n",
+    "    padding = int(onp.around(spec.width / spec.grid_spacing)) - design.shape[0]\n",
+    "    padded_design = onp.pad(design, ((padding // 2, padding // 2), (0, 0)), mode=\"edge\")\n",
+    "    params = dataclasses.replace(dummy_params, array=padded_design)\n",
+    "\n",
+    "    response, aux = challenge.component.response(params)\n",
+    "\n",
+    "    # Compute the length scale in nanometers for the binary design.\n",
+    "    length_scale = (\n",
+    "        onp.amin(imageruler.minimum_length_scale(design > 0.5)) * grid_spacing * 1000\n",
+    "    )\n",
+    "\n",
+    "    if \"Mo\" in fname:\n",
+    "        key = \"mo\"\n",
+    "    elif \"Rasmus\" in fname:\n",
+    "        key = \"rasmus\"\n",
+    "    elif \"invrsio\" in fname:\n",
+    "        key = \"invrsio\"\n",
+    "\n",
+    "    results[key].append(\n",
+    "        {\"response\": response, \"length_scale\": length_scale, \"aux\": aux}\n",
+    "    )"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "Plotting, we see results very similar to those of figure 3 of \"[Validation and characterization of algorithms and software for photonics inverse design](https://opg.optica.org/josab/abstract.cfm?URI=josab-41-2-A161)\" by Chen et al.\n",
+    "\n",
+    "Differences are explained by updates to the imageruler algorithm, and the fact that we are using a different simulation algorithm to model the metalenses."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "ax = plt.subplot(111)\n",
+    "\n",
+    "for key, result in results.items():\n",
+    "    enhancement = onp.asarray([onp.mean(r[\"response\"].enhancement_ex) for r in result])\n",
+    "    length_scale = onp.asarray([onp.amin(r[\"length_scale\"]) for r in result])\n",
+    "    order = onp.argsort(length_scale)\n",
+    "    ax.plot(length_scale[order], enhancement[order], \"o-\", lw=3, ms=10, label=key)\n",
+    "\n",
+    "ax.set_xlabel(\"Length scale (nm)\")\n",
+    "ax.set_ylabel(\"Wavelength average intensity at focus (a.u.)\")\n",
+    "ax.set_ylim((5, 25))\n",
+    "_ = plt.legend()"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
-   "source": []
+   "source": [
+    "## Metalens optimization\n",
+    "\n",
+    "To optimize a metalens, you may follow the recipe in the `ceviche_challenge` and `metagrating_challenge` notebooks."
+   ]
   }
  ],
  "metadata": {