Polarization problem

Birefringence in nominally symmetric fibers

In principle, a fiber with a perfectly rotationally symmetric design should have no birefringence. Therefore, it should fully preserve the polarization of the light. In practice, however, a certain amount of birefringence is always caused by imperfections in the fiber (eg, slight ellipticity of the fiber core) or bends. As a result, the polarization state of the light changes over relatively short lengths of fiber—sometimes just a few meters, sometimes much faster.

Note that the difference in refractive index between polarization directions in fiber is not necessarily greater than in other devices. However, fibers tend to be very long, so even a small difference in index can have a big impact.

Another important aspect is that the resulting polarization changes are not only random and unpredictable, but also strongly dependent on the wavelength, the temperature of the fiber along its entire length, and any bending of the fiber. Therefore, adjusting the polarization state is usually not very helpful, e.g. using a fiber optic polarization controller (see below); some small changes in environmental parameters or wavelength may again destroy the polarization.

Fiber Polarization Controller

Strong bending of the fiber introduces birefringence. This means that some suitable length of fiber bent at a certain radius and fixed on a coil can have a relative phase delay of π or π/2, for example, between two polarization directions. Therefore, it can act like a λ/2 wave plate (half wave plate) or a λ/4 wave plate (quarter wave plate). If the entire coil is rotated about an axis coincident with the input and output fibers, an effect similar to that of rotating a bulk waveplate in a free-space laser beam is obtained. One often uses a combination of an effective quarter-wave plate coil with a half-wave plate coil and another quarter-wave plate coil in series to convert some input polarization state to any desired polarization state. Such a fiber-optic polarization controller (Fig. 1) can operate over a considerable wavelength range.

Figure 1: The “bat ears” polarization controller consists of three fiber optic coils that rotate about the axis of the input fiber.

As mentioned earlier, the problem may still remain that the input polarization state will drift with changing environmental conditions, so the fiber polarization controller must be frequently readjusted to maintain a constant output polarization state.

PM fiber

Optical fibers can be made into PM fibers (PM fibers) – but not avoiding any birefringence! Instead, significant birefringence is intentionally introduced. Such fibers are thus highly birefringent fibers (HIBI fibers).

Basically there are two common ways to do this:

• Optical fibers can be made with elliptical cores. This causes some degree of shape birefringence. Of course, the fiber mode is also affected by the elliptical shape, and the optical coupling efficiency to the circular core fiber is reduced.
• Some mechanical stress can be applied, for example by introducing stress rods made of different glasses. See Figure 2 for some typical implementations.

Figure 2: Polarization maintaining PANDA fiber (left) and bowtie fiber (right). Built-in stress elements made of different types of glass are shown in dark gray tones.

Note: PM fibers do not retain any polarization state of the injected light! It only does this for linearly polarized light, where the polarization direction must be one of two orthogonal directions, such as along the line between the stress rods or perpendicular to it. The value of β for certain wavelengths will depend significantly on the polarization direction.

What happens if we inject monochromatics with other linear polarization directions? This can be considered as a superposition of two fundamental polarization states. After a short time of propagation, these components will acquire significantly different phase delays (due to their different β values). Therefore, they will no longer bind to the original linear polarization state, but generally to some elliptical state. However, after an integer multiple of the polarization beat length, the linear polarization is again obtained.

For non-monochromatic light, the situation becomes more complicated. Over a certain length of fiber, different wavelength components will experience different polarization-dependent phase shifts, so the resulting polarization state becomes wavelength-dependent. Converting it back to a linear state would be a difficult task — a simple polarization controller can’t do it.

The need to align the input polarization state with the fiber axis to maintain the polarization state is of course a serious practical disadvantage of polarization maintaining fibers. Fabrication of PM fiber optic devices requires more work and additional equipment is required for this. Also, not all fiber optic assemblies are available as PM versions. On the other hand, adverse effects of drifting polarization state can be safely avoided by PM setting, which might otherwise require other measures.

Note that the introduced birefringence essentially removes any effect of some small additional random birefringence, which might be caused by, for example, modest bending. This random effect may only alter the local polarization very slightly, but usually does not have any significant effect over longer lengths. One can understand this by considering mode coupling: significant mode coupling requires a perturbation with a period equal to the beat period of the two polarization states. For strong birefringence, this beat (polarization beat length) period is rather short (e.g., a few millimeters), and the usual perturbations are too “slow” spatially to cause any significant coupling, or at least according to which the polarization beat does not have Strong spatial Fourier component.

Polarization insensitive design

Another way to eliminate polarization problems is to design the device so that polarization does not matter. For example, this method is commonly used in fiber optic communications. Just be careful not to use components that could cause a lot of polarization dependent loss or be dependent on a particular polarization state. For example, electro-optic modulators typically cannot be used and require careful design of any semiconductor device with low polarization dependence. Some polarization effects still exist, which can limit the performance of extremely fast fiber links. In particular there is the phenomenon of polarization mode dispersion (PMD), which can be quantified as differential group delay (DGD): signal components with different polarizations can take slightly different times to travel through a fiber optic cable, which can degrade signal quality. However, for short transmission distances and/or moderate bit rates, PMD is not as much of an issue.