Polarization extinction ratio (PER) is a measure of how cleanly a fiber, waveguide, or optical component maintains light in a single polarization state. Expressed in decibels, it compares the power in the desired polarization mode to the power that has leaked into the unwanted orthogonal mode. A higher number means better polarization purity. The metric sounds straightforward, but its practical measurement turns out to be surprisingly tricky, with strong wavelength dependence and several common techniques that can give misleading results.
What the Number Actually Tells You
Think of polarized light as vibrating in a single, well-defined direction. In a perfect world, every photon coming out of a polarization-maintaining fiber or a polarizing beam splitter would vibrate in exactly the orientation you intended. In reality, some fraction of the light ends up in the perpendicular direction due to imperfections, bends, splices, or material limitations. PER quantifies how small that leakage fraction is.
Because the power difference between the wanted and unwanted polarization modes can span many orders of magnitude, engineers express PER in decibels rather than as a simple ratio. A PER of 20 dB means the desired polarization is a hundred times stronger than the leakage. At 30 dB that ratio jumps to a thousand to one, and at 40 dB it reaches ten thousand to one. For many standard fiber-optic sensors and telecom components, a PER somewhere in the 20 to 30 dB range is considered good. Cutting-edge integrated photonic devices have pushed well beyond that, as we’ll see below.
PER is sometimes called “polarization crosstalk,” and the two terms appear interchangeably in datasheets and journal papers. One well-known study on fiber-based devices explicitly notes that the polarization extinction ratio is “also called polarization cross talk,” and the paper devotes considerable effort to clarifying the definition because the two names can cause confusion about sign conventions and whether the ratio refers to a device or to the light emerging from it.1Applied Optics. Definition, meaning, and measurement of the polarization extinction ratio of fiber-based devices
Why Wavelength Matters So Much
One of the least intuitive things about PER is that it can change dramatically with wavelength. Even for something as simple as a single fiber patchcord, the extinction ratio at one wavelength can be quite different from the extinction ratio a few tens of nanometers away. That same clarifying study on fiber-based devices highlights this strong wavelength dependence and warns that broadband or “white” light sources should not be used for most PER measurements, precisely because the wavelength averaging they introduce obscures the true polarization behavior at any given wavelength.2Applied Optics. Definition, meaning, and measurement of the polarization extinction ratio of fiber-based devices
This wavelength sensitivity carries over into integrated photonic splitters designed to separate the two polarization modes on a chip. A compact polarization beam splitter built from subwavelength grating couplers, for example, achieved an extinction ratio of about 32 dB for one polarization mode and roughly 21 dB for the other at the standard telecom wavelength of 1550 nm. But its usable bandwidth, defined as the spectral window in which the extinction ratio stays above 18 dB, was only about 81 nm.3Optics Letters. Compact and high extinction ratio polarization beam splitter using subwavelength grating couplers Step outside that window and the polarization purity drops. A different design using tilted subwavelength gratings reported a peak ER of about 34 dB for the TM mode, also centered near 1550 nm, with the bandwidth for keeping ER above 20 dB reaching roughly 75 nm on one polarization and a wider 133 nm on the other when an additional waveguide structure was added at the output.4Applied Optics. Broadband and high-extinction-ratio polarization beam splitter on tilted subwavelength gratings waveguides
The takeaway for anyone specifying or purchasing a polarization-sensitive component is that quoting a single PER number without also specifying the wavelength (or wavelength range) is incomplete. A device rated at 30 dB PER at 1550 nm might offer only 15 dB at 1530 nm. Always check the spectral window, not just the peak specification.
Measuring PER Correctly
Measuring polarization extinction ratio sounds like it should be simple: put polarized light in, rotate an analyzer at the output, and record the maximum and minimum power. Divide one by the other and convert to decibels. In practice, this “rotating polarizer” method has well-documented weaknesses. The same Applied Optics paper that clarified the definition also explains that using a rotating polarizer with a monochromatic source can yield inaccurate results, and states that a polarimeter (an instrument that fully characterizes the polarization state) is needed if accuracy matters.5Applied Optics. Definition, meaning, and measurement of the polarization extinction ratio of fiber-based devices
Why does the rotating-polarizer approach fall short? The method assumes you can align the analyzer perfectly with the two polarization axes and that nothing else in the setup is drifting or adding its own birefringence. In fiber systems, slight bends between the device under test and the rotating analyzer can shift the polarization state, and the measurement reflects the whole optical path rather than just the device. A full polarimetric measurement captures the complete polarization state in one shot, eliminating much of this ambiguity.
For situations where you need to know how PER varies along the length of a waveguide rather than just at the output, optical low-coherence interferometry offers a distributed measurement. This technique works by exploiting the interference between the light in the intended polarization and the light that has coupled into the orthogonal mode; because the two modes travel at different speeds in a birefringent waveguide, the interference pattern reveals where along the waveguide the coupling events occur.6Journal of Lightwave Technology. Polarization Crosstalk Dependence on Length in Silica-Based Waveguides Measured by Using Optical Low Coherence Interference That kind of spatial mapping is especially useful in manufacturing, where you want to find and fix the specific splice or bend that is degrading polarization purity.
Polarization Extinction in Fibers
Polarization-maintaining (PM) fibers are engineered with built-in stress rods or elliptical cladding that create two distinct polarization axes. Light launched along one axis is supposed to stay there. But external stresses, bends, and splices can transfer energy to the other axis, degrading PER. The longer the fiber run, the more opportunities for such coupling.
One mechanism that has been studied in detail is periodic microbending, where repeated small lateral forces act on a PM fiber at regular intervals. Researchers have shown that the mode coupling this induces is strongly polarization dependent and that the coupling strength depends on the angle between the fiber’s birefringence axes and the direction of the bending force.7Optics Letters. Periodic microbending-induced core-to-cladding mode coupling in polarization-maintaining fibers In practical terms, this means that how you mount, route, or clamp a PM fiber cable matters. A fiber squeezed against a grooved surface at just the wrong angle relative to its stress rods will lose polarization purity far faster than one oriented favorably.
Temperature changes add another complication. As the fiber heats or cools, differential thermal expansion between the stress elements and the surrounding silica alters the internal stress profile, shifting the birefringence slightly. In laboratory setups where PER of 30 dB or better is needed, temperature-controlling the fiber or at least minimizing thermal cycling is common practice. Field deployments rarely have that luxury, which is one reason why telecom and sensor systems build in margin above the minimum required PER.
Polarization Beam Splitters on a Chip
In integrated photonics, where entire optical circuits sit on a silicon chip a few millimeters across, one of the most critical building blocks is the polarization beam splitter (PBS). This is the on-chip equivalent of a free-space Wollaston prism: it takes incoming light carrying both polarization modes and routes each to a separate output waveguide. The extinction ratio of the PBS directly limits how well the rest of the circuit can process each polarization independently.
The compact grating-based splitters mentioned earlier represent one design family, achieving extinction ratios in the 20 to 34 dB range depending on the polarization mode and waveguide geometry.8Applied Optics. Broadband and high-extinction-ratio polarization beam splitter on tilted subwavelength gratings waveguides But a more aggressive approach uses anisotropic metamaterial claddings around the waveguide to engineer enormous birefringence, so the two polarization modes separate extremely quickly. One such device demonstrated extinction ratios of 58 to 60 dB for the TE mode and 46 to 48 dB for the TM mode, all in a coupling length of only 14.5 micrometers.9OSA Advanced Photonics Congress 2021. Ultra-high extinction-ratio silicon photonic polarization beam splitter with anisotropic metamaterials Those numbers are strikingly high. An extinction ratio of 60 dB means the unwanted polarization mode is a million times weaker than the desired one. For reference, most commercial off-the-shelf fiber PBS components sit in the 20 to 30 dB range, so a 60 dB on-chip device represents a huge jump.
These metamaterial designs are still largely in the research stage, and fabrication tolerances at the sub-wavelength scale can erode the real-world performance compared to the simulated ideal. Still, the direction is clear: on-chip polarization management is getting dramatically better, and extinction ratios that were once considered impossible on silicon are becoming plausible.
How PER Affects Telecommunications
Modern high-capacity fiber-optic links frequently use polarization-division multiplexing, where two independent data streams are transmitted simultaneously on the same wavelength by assigning each to a different polarization state. At the receiver, a coherent detection scheme separates the two streams. If the polarization extinction ratio of the system’s components is poor, some of one stream bleeds into the other. This is cross-polarization crosstalk, and it directly increases the bit error rate.
A simulation study of a 100 Gbps coherent system using an advanced modulation format over 100 km of standard single-mode fiber found that cross-polarization effects degrade the bit error rate performance in a predictable way: the greater the misalignment angle between the signal’s polarization and the receiver’s reference, the larger the power penalty needed to achieve a given error rate.10Results in Optics. Cross-polarization induced crosstalk impact analysis on the BER performance of 100-Gbps CPDM 8-QAM CO-FOC system over unrepeatered 100-km SSMF link In other words, every dB of extinction ratio you lose at the polarization-handling components translates into a real penalty in data throughput or distance reach. For backbone networks pushing hundreds of gigabits per second per channel, even a few dB of extra polarization crosstalk can be the difference between a link that works and one that doesn’t meet its error-rate target.
This is why telecom equipment makers specify PER tightly for components like polarization beam combiners, rotators, and the modulators themselves. A modulator that accepts two polarization tributaries and combines them for transmission might need a PER well above 20 dB to keep the crosstalk penalty within acceptable bounds. Couplers and connectors in the signal path each contribute their own small crosstalk, and these contributions add up.
Dynamic Polarization Control
In many real-world systems, the polarization state of light arriving at a receiver is not constant. Wind-induced vibration of aerial fiber cables, temperature swings underground, and even tiny mechanical disturbances inside equipment racks all cause the polarization to wander over time. A static high-PER component at the transmitter doesn’t help if the polarization has scrambled by the time it reaches the receiver.
Dynamic polarization controllers address this by actively tracking and correcting the incoming polarization state in real time. One recent demonstration on a silicon photonics chip used a feedback algorithm to continuously adjust on-chip phase shifters, achieving a dynamic polarization extinction ratio greater than 25 dB.11Chinese Optics Letters. Silicon photonics integrated dynamic polarization controller Integrating this function onto the same chip as the rest of the receiver circuitry is attractive because it shrinks the footprint and potentially lowers cost compared to standalone polarization controllers that use bulky piezo-driven fiber squeezers or lithium niobate crystals.
The 25 dB figure achieved dynamically is lower than the 58 to 60 dB demonstrated in the metamaterial-based beam splitter described above, but the comparison is somewhat apples-to-oranges. The beam splitter is a passive device measured under controlled conditions, while the dynamic controller is tracking a moving target and re-optimizing on the fly. For a real-time feedback system, 25 dB of maintained extinction is a solid result that meets the requirements of most coherent telecom receivers.
Real-World Polarizers and Imperfect Extinction
The concept of extinction ratio also applies to the polarizers themselves, not just to fibers and waveguides. No physical polarizer transmits only one polarization state with absolutely zero leakage of the other. High-quality Glan-Thompson prisms used in laboratory setups can reach extinction ratios above 50 dB, while typical sheet-film polarizers used in everyday applications might sit closer to 20 to 30 dB. In polarimetry systems, the finite extinction ratio of the polarizer feeds directly into measurement error.
A study on extracting optical rotation angles in turbid media showed that the non-ideal extinction ratio of the rotating polarizer had to be explicitly modeled to get accurate results. The Mueller matrix representation of the polarizer was modified to include the leakage term, rather than treating the polarizer as ideal.12Measurement. Extraction of optical rotation angle in a turbid media using polarimetry system with single rotating polarizer The practical lesson here is that if you’re building a system that depends on clean polarization, you need to account for the imperfect extinction of every element in the optical path, including the polarizers you’re using to characterize the rest of the system.
This creates a measurement chicken-and-egg problem that frustrates newcomers to the field. You want to measure the PER of your fiber or waveguide, but your measurement setup includes polarizers and analyzers that have their own finite extinction ratios. If the device under test has better extinction than your measurement gear, you’re measuring the setup rather than the device. This is another reason why full polarimetric approaches, which avoid relying on a single analyzer’s purity, tend to give more trustworthy results for high-PER components.
Where the Specifications Fail You
Datasheet PER values for fiber-optic components are typically measured at a single wavelength, at a fixed temperature, with the fiber in a carefully controlled routing configuration. The moment you splice the component into a real system with different routing, temperature extremes, or a different operating wavelength, the actual PER can differ considerably from the spec sheet. This gap between laboratory performance and field performance is one of the most common sources of frustration in polarization-sensitive systems.
Connector alignment is another frequently underestimated factor. Polarization-maintaining fiber connectors must be rotationally aligned so that the fiber’s slow axis matches the mating component. Misalignment of even a few degrees can reduce the effective PER by several dB. Some connector types include alignment keys or ferrule marks to minimize this, but the alignment is only as good as the original marking accuracy and the mechanical tolerance of the key slot.
Cascading effects compound the problem. If a system has ten PM fiber connections in series, and each contributes a small polarization error, the cumulative extinction ratio at the end of the chain can be much worse than any single component’s spec would suggest. System designers typically budget a PER margin for each connection and splice, then calculate the worst-case total. Being too optimistic at this stage is a reliable recipe for a system that tests beautifully on the bench and fails in the field.
Applications Beyond Telecom
Fiber-optic gyroscopes, used in aircraft and spacecraft navigation, are among the most PER-sensitive applications outside of telecommunications. These sensors detect rotation by measuring tiny phase differences between two counter-propagating beams in a coil of PM fiber. Any polarization crosstalk within the coil introduces a spurious signal that looks like rotation, degrading the gyroscope’s accuracy. High-performance fiber gyroscopes require PER values above 30 dB throughout the sensing coil, and achieving that over hundreds of meters of coiled fiber while the coil is subject to vibration and temperature changes is a serious engineering challenge.
Coherent lidar systems, quantum key distribution setups, and polarization-resolved spectroscopy all have their own PER budgets that must be managed. In quantum communications, the polarization state often encodes the quantum information directly, so even tiny amounts of crosstalk can compromise the security guarantees of the protocol. These systems tend to use the highest-quality PM fibers and connectors available, and often include in-line polarization monitoring to catch degradation before it causes errors.
Biomedical imaging presents yet another context. Polarization-sensitive optical coherence tomography, for example, uses polarization contrast to distinguish tissue types. The extinction ratio of the fiber and optics in the probe determines the minimum birefringence difference the system can detect. A lower PER blurs the polarization contrast and makes it harder to differentiate, say, collagen-rich tissue from surrounding structures. Researchers in this space don’t always use the term “PER” explicitly, but the underlying physics is the same: how well can you keep polarization states separate through the instrument?