adding anisotropic eme faq

_faqs/how-do-i-simulate-a-bent-anisotropic-waveguide-with-eme.md

@@ -0,0 +1,42 @@
+---
+title: How Do I Simulate a Bent Anisotropic Waveguide with EME?
+date: 2026-04-23 11:26:28
+enabled: true
+category: "EME"
+---
+# How do I simulate a bent anisotropic waveguide with EME?
+
+Unlike isotropic bends, which can be modeled with a single EME cell (see [how can I simulate bent waveguides with EME](https://www.flexcompute.com/tidy3d/faq/docs/faq/how-can-i-simulate-bent-waveguides-with-eme.html){: .color-primary-hover}), anisotropic bends require **multiple EME cells** and a convergence sweep over `num_cells` in [tidy3d.EMEUniformGrid](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.EMEUniformGrid.html){: .color-primary-hover}.
+
+The reason is that in an anisotropic medium — defined with [tidy3d.FullyAnisotropicMedium](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.FullyAnisotropicMedium.html){: .color-primary-hover} or [tidy3d.AnisotropicMedium](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.AnisotropicMedium.html){: .color-primary-hover} — the permittivity tensor is fixed in the lab frame, but the propagation direction rotates along the arc. The local eigenmodes therefore change continuously, producing polarization mixing and back-reflections. A single EME cell misses this effect and **overestimates transmission**. Dividing the arc into several cells in [tidy3d.EMESimulation](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.EMESimulation.html){: .color-primary-hover} — each with its own curved eigenmode set defined via `bend_radius` in [tidy3d.EMEModeSpec](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.EMEModeSpec.html){: .color-primary-hover} — recovers the correct result.
+
+
+
+<div markdown class="code-snippet">
+{% highlight python %}
+python
+
+mode_spec = td.EMEModeSpec(
+    num_modes=num_modes,
+    target_neff=target_neff,
+    bend_radius=radius,
+    bend_axis=1,
+)
+
+eme_sim = td.EMESimulation(
+    size=(np.pi * radius, plane_size[0], plane_size[1]),  # arc length along propagation axis
+    center=(0, 0, 0),
+    structures=[waveguide],
+    medium=background_medium,
+    axis=0,
+    freqs=[freq0],
+    eme_grid_spec=td.EMEUniformGrid(num_cells=num_cells, mode_spec=mode_spec),
+    grid_spec=grid_spec,
+    store_coeffs=True,
+)
+{% endhighlight %}
+{% include copy-button.html %}</div>
+
+
+For a complete example featuring a 180° LiNbO₃ bend, an FDTD reference simulation, and a cell-count convergence study, refer to [this tutorial](https://www.flexcompute.com/tidy3d/examples/notebooks/AnisotropicBendsEME/){: .color-primary-hover}.
+

docs/eme.rst

@@ -9,3 +9,4 @@ EME
    faq/how-can-i-simulate-bent-waveguides-with-eme.md
    faq/how-do-i-set-an-eme-simulation.md
    faq/what-is-eme.md
+   faq/how-do-i-simulate-a-bent-anisotropic-waveguide-with-eme.md

docs/faq/how-do-i-simulate-a-bent-anisotropic-waveguide-with-eme.md

@@ -0,0 +1,45 @@
+# How Do I Simulate a Bent Anisotropic Waveguide with EME?
+
+| Date       | Category    |
+|------------|-------------|
+| 2026-04-23 11:26:28 | EME |
+
+
+# How do I simulate a bent anisotropic waveguide with EME?
+
+Unlike isotropic bends, which can be modeled with a single EME cell (see [how can I simulate bent waveguides with EME](https://www.flexcompute.com/tidy3d/faq/docs/faq/how-can-i-simulate-bent-waveguides-with-eme.html)), anisotropic bends require **multiple EME cells** and a convergence sweep over `num_cells` in [tidy3d.EMEUniformGrid](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.EMEUniformGrid.html).
+
+The reason is that in an anisotropic medium — defined with [tidy3d.FullyAnisotropicMedium](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.FullyAnisotropicMedium.html) or [tidy3d.AnisotropicMedium](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.AnisotropicMedium.html) — the permittivity tensor is fixed in the lab frame, but the propagation direction rotates along the arc. The local eigenmodes therefore change continuously, producing polarization mixing and back-reflections. A single EME cell misses this effect and **overestimates transmission**. Dividing the arc into several cells in [tidy3d.EMESimulation](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.EMESimulation.html) — each with its own curved eigenmode set defined via `bend_radius` in [tidy3d.EMEModeSpec](https://docs.flexcompute.com/projects/tidy3d/en/latest/api/_autosummary/tidy3d.EMEModeSpec.html) — recovers the correct result.
+
+
+
+
+
+```python
+python
+
+mode_spec = td.EMEModeSpec(
+    num_modes=num_modes,
+    target_neff=target_neff,
+    bend_radius=radius,
+    bend_axis=1,
+)
+
+eme_sim = td.EMESimulation(
+    size=(np.pi * radius, plane_size[0], plane_size[1]),  # arc length along propagation axis
+    center=(0, 0, 0),
+    structures=[waveguide],
+    medium=background_medium,
+    axis=0,
+    freqs=[freq0],
+    eme_grid_spec=td.EMEUniformGrid(num_cells=num_cells, mode_spec=mode_spec),
+    grid_spec=grid_spec,
+    store_coeffs=True,
+)
+```
+
+
+
+
+For a complete example featuring a 180° LiNbO₃ bend, an FDTD reference simulation, and a cell-count convergence study, refer to [this tutorial](https://www.flexcompute.com/tidy3d/examples/notebooks/AnisotropicBendsEME/).
+

faq_categories.json

@@ -316,7 +316,8 @@
       "_faqs/how-to-define-the-number-of-modes-for-an-eme-simulation.md",
       "_faqs/how-can-i-simulate-bent-waveguides-with-eme.md",
       "_faqs/how-do-i-set-an-eme-simulation.md",
-      "_faqs/what-is-eme.md"
+      "_faqs/what-is-eme.md",
+      "_faqs/how-do-i-simulate-a-bent-anisotropic-waveguide-with-eme.md"
     ]
   },
   {
@@ -375,4 +376,4 @@
       "_faqs/how-do-i-loop-through-tidy3d-web-batchdata-without-loading-all-of-the-data-into-memory.md"
     ]
   }
-]
+]