loss in PCF


I have the problem in loss.
I am using my own material index values for core and cladding (silica clad). Is it required to fit the material using “fit material with coefficient model” or not. if so why we have to do the procedure. In the following pcf https://kb.lumerical.com/en/diffractive_optics_pcfibre.html they choose without material fitting using corning silica, and the results explained according to that (bending the fiber). the figure bend radius vs loss is decreasing is satisfied.
(I will take mesh cells more like 500, and pml conditions like LMA fiber optimized conditions)
give me exact notation I have to consider the fitting model or not.?

thanks in advance


Hi @Nanda,

If you are only interested in getting the bend loss versus bending radius then it’s not important to fit the material using the multi-coefficient model.

The material fitting becomes important if you want to get dispersion results or perform frequency sweeps since if you don’t use the multi-coefficient model to generate a smooth fit of the material data over frequency, this will lead to discontinuties in the calculated dispersion as shown in the following example:


hi nlui
thank you for your reply
I am interested in loss not dispersion. If I am using without fitting the loss values are coming in negative for example -2.33E-6. kindly clarify the point can I take the value as positive.
an other thing I need the loss for range of wavelengths, in that case i will calculate each wavelength i will not sweep the frequency. I will take care that point, is it good ?

kindly let me your valuable comments
thanks in advance


Hi @Nanda,

If you would like to plot any value over wavelength, then I would recommend using the multi-coefficient model to generate a fit of the material data over the wavelength range of interest. This would result in a smoother plot of loss over wavelength, and running the simulations with the material fitting enabled does not require additional memory or time so I would recommend it.

We typically want to use the material fit for these types of plots instead of using linear interpolation of the material data at each wavelength since the measured material data has some measurement error which makes the results bumpy.

A very small negative loss value could be error introduced by using PML boundary conditions. It may be the case that with the bending radius that you are using, the radiative loss due to bending is negligible and the error introduced by the PML boundaries is larger. I would expect that if you continue to make the bending radius smaller, you will start seeing the radiative loss. The error from the PML boundary conditions is discussed on the following page:

Hopefully this helps!


hi nlui
thank you firstly,

at first if I am using fitting for range of wavelengths the loss is decreasing in 4 order magntiude (for example without the loss is 1e-10, after fitting it is 1e-6).
and next about the negative sign I have seen your kb page it mentioned negative value is close to zero. in our case (fibers) the loss is usually 1e-10 dB/cm in km it is very low loss. I cant take zero approximation. kindly let me some clarification
[I am attaching my personal file to your mail and in that I have defined clad as silica from material data base and core is my own material, to cover clad we have used the air medium. I have calculated loss for core and clad mode. The core mode is positive loss but clad mode is negative loss we need clad mode also. To solve this I increased PML distance from the clad even in that case also clad mode is negative and I increased pml values from kappa 2 and sigma 10 in that case also it is negative. **Finally when I am using fitting material condition the loss vlaues coming positive but the order of loss value is decreasing like as mentioned above**]
with advance thanks



I have received your email with file. I’ll look into your file and reply to you directly.