Optical isolators are widely used in optical fiber communication systems, optical fiber sensing systems and fiber lasers. The basic and common principle for optical isolators is Faraday effect. However, the device structures and characteristics are variable, which are detailed as follow.
Free-space Optical Isolator
The structure of a free-space optical isolator is shown in Fig.1, which comprises two polarizers, a Faraday rotator (FR) and a magnet ring. The transmission axes of the two polarizers are aligned with 45º angle and the FR has a fixed rotatory angle of 45º in a saturated magnetic field.
For the forward light(linearly polarized (LP)),itfirst passes polarizer #1 and then rotated by the FR. It passes polarizer #2 with little loss finally because its optical vector is aligned to the transmission axis. For the backward light, it first passes polarizer #2 and then rotated by the FR. It is blocked by polarizer #2 finally because its optical vector is perpendicular to the transmission axis. Thus the forward light is let pass, while the backward light is isolated.
Free-space isolator is characterized by simple structure and low cost. However, it is polarization dependent, which limits its applications. It requires the forward light to be LP polarized and the optical vector aligned to the first polarizer. Thus it can’t be employed in-line (optical fiber)because the SOP (state of polarization) of optical signal in optical fiber is random. The major application of the free-space isolators is inlaser diodes (LDs).
A FP LD (Fabry-Perot laser diode) emits randomly polarized light, while DFB LD (distributed feedback laser diode) emits LP light. Comparing to FP LDs, DFB LDs have better monochromaticity, higher output power and higher modulation rate, which enable its more applications. In an optical transmitter, a free-space isolator is placed between the DFB LD chip and optical fiber. Any back reflection light from the fiber line is isolated to prevent damage on the LD chip.
BD-type Optical Isolator
In-line applications, such as EDFA, need polarization independent optical isolators whichcan let pass forward light with any SOP. The first polarization independent optical isolator is based on beam displacers (BDs), as shown in Fig.2. It consists two BDs, a FR, a HWP (half waveplate), a magnet ring and two fiber collimators.
The forward light from the input collimator incidents on the first BD. It is laterally separated as o-ray and e-ray in the BD block. The two rays pass the FR+HWP component and the SOPs are transformed as e→o and o→e. The two rays are recombined by the second BD block and finally received by the output collimator. Thus the forward light is let pass.
The backward light from the output collimator is first laterally separated by the second BD and then transformed by the FR+HWP component. The two rays are laterally separated more, instead of being recombined. The two separated rays deviate from the center of the input collimator and can’t be received by it. Thus the backward light is isolated.
The BD-based isolator can let pass forward light with any SOPwith two BDs for polarization diversity. Thus it is called polarization independent optical isolator.
Note that the input and output fiber collimators are not coaxial due to asymmetric separation of rays by the BD blocks, which is not convenient for assembly and package of the device. Actually, coaxial design can be realized with the optimized BDs we formally mentioned.
The backward light is isolated due to the lateral displacement of rays in the first BD block. Therelation between isolation and lateral displacement of rays is calculated and shown as Fig.3. The rays need to be separated by 0.6mm in the BD blocks to obtain an isolation of 40dB. Uniaxial crystal YVO4 is widely used to make BD blocks due to its high birefringence. The maximum deviation ratio between o-ray and e-ray is d:L=1:10 (d is the lateral separation between the two rays and L is the length of the BD block), which means that the minimum BD length is 6mm. The size is rather big and YVO4 is expensive. Thus BD-type optical isolators are rarely used in optical fiber communication systems, which rate the most demand for optical isolators.
Wedge-type Optical Isolator
The structure of a wedge-type optical isolator is shown in Fig.4. It includes two fiber collimators and an isolator core. The core structure is shown in Fig.5, including two birefringent wedges, one 45º FR and a magnet ring.
The optical axes of the wedges are both parallel to the right angle surfaces, while the orientations have an intersection angle of 45º, as shown in Fig.5. The forward natural light from the first collimator is divided as o-ray and e-ray in wedge #1. Then the optical vectors are rotated by 45º after passing the FR. When the two rays enter wedge #2, they keep as o-ray and e-ray. Thus the SOP transformation of the two rays in the two wedges is o→o and e→e, respectively. The refractive index keeps unchanged for each ray. The whole isolator core acts as a parallel plate for the forward light. The rays pass the plate obliquely and keep parallel to each other. Each ray has a lateral offset and the walk-off (difference between the offsets of the two rays) is small (~10μm). The two rays are received by the second collimator with little loss.
For the backward light, the SOP transformation of the two rays in the two wedges is o→e and e→o, respectively. The isolator core functions as a Wollaston prism for the backward light. The two rays are both angularly deviated and can’t be received by the input collimator. The relation between isolation and angular deviation of the rays is shown in Fig.6. Isolation increase rapidly with the increment of angular deviation. An angular deviation of 0.5º is enough to isolate the backward light by 50dB.
Actually, the isolation of a wedge-type optical isolator is limited by extinction ratio (ER) of the FR. The commercial FR has a typical ER of 40dB, which means that 1/10000 optical power is detected with the transmission axis of a polarizer aligned to y axis when the optical vector of a LP light is rotated to x axis. The ER of the FR limits the highest isolation to be ~40dB.
Dual-stage Optical Isolator
In some special applications that require higher isolation, dual-stage optical isolators are demanded. There are three typical structures of dual-stage wedge-type optical isolator.
Fig.7 shows structure #1 of the dual-stage isolator core, which consists two isolator cores relatively rotated by 45º. Thus forthe forward light, the SOP transformation of the two rays in the four wedges is o→o→e→e and e→e→o→o, respectively. The dual-stage isolator core acts as two parallel plates for the forward light, although the principle sections of the two plates have an intersectional angle of 45º. The two rays pass the plates obliquely and keep parallel to each other. Each ray has a lateral offset and the walk-off is small (~15μm). The two rays are received by the output collimator with little loss.
For the backward light, the SOP transformation of the two rays in the four wedges is o→e→o→e and e→o→e→o, respectively.The isolator core acts as two Wollaston prisms. The two rays are angularly deviated more than that in a single-stage isolator. Actually, the improvement of isolation is not attributed to more angular deviation of the backward light. It owes to one more FR to double the ER.
The dual-stage optical isolator in Fig.7 can obtain high isolation. However, the assembly efficiency and yield rate are not good because the relative 45º rotation between two isolator cores always introducesunexpected scattering on the backward light and degrades the isolation of the device.
The second solution for dual-stage optical isolator is shown in Fig.8. Four birefringent wedges and two FRs are aligned in series. The optical axes of the wedges are also shown in the figure. Note that the intersection angles between P1 and P2, P2 and P3, P3 and P4 are 45º, 90º and 45º, respectively. Both the FRs rotate the optical vector counter-clockwise by 45º (with the observer facing the forward light) .
For the forward light, the SOP transformation of the two rays in the four wedges is o→o→e→e and e→e→o→o, respectively. The dual-stage isolator core acts as two parallel plates with principle sections parallel to each other.The two rays pass the plates obliquely and keep parallel to each other. Each ray has a lateral offset and the walk-off is small (~20μm). The two rays are received by the output collimator with little loss.
For the backward light, the SOP transformation of the two rays in the four wedges is o→e→o→e and e→o→e→o, respectively.The isolator core acts as two Wollaston prisms. The two rays are angularly deviated more than that in a single-stage isolator. As we know, the improvement of isolation owes to one more FR to double the ER.
Let’s examine the structure in Fig.8. If the optical axes of wedges P2 and P3 are not strictly perpendicular to each other, what happens? The backward light is first separated into two rays in wedge P4 and then further separated into four rays in P2. Among the four rays, two have relatively higher intensity and their SOP transformation in the four wedges is o→e→o→e and e→o→e→o, respectively. These two rays are blocked because of angular deviation. The other two rays have rather lower intensity and their SOP transformation in the four wedges is o→e→e→o and e→o→o→e, respectively. The SOP transformation means that wedges P2 and P3 can be treated as a parallel plate, while wedges P1 and P4 can be treated as another parallel plate. The latter two rays pass two parallel plates. They keep the transmission direction and are finally received by the input collimator. Thus the isolation is degraded.
Although the intensity of the latter two rays is rather low. The influence on the isolation is significant. The intensity of the latter two rays is sin2δ given the intersection angle between the axes of P2 and P3 as 90º±δ. Thus the relation between the isolation and assembly error δ is obtained as Fig.9. The isolation is obtained as 35dB or 29dB when δ is 2º or 1º. In order to obtain an isolation of 55dB expected for a dual-stage isolator, the assembly error δ is required to be <0.1º, which is quite difficult for production.
Based on analysis of above structure #2, a revised structure for dual-stage optical isolator is provided as Fig.10. The optical axes orientations are the same as those in Fig.8. The only difference is that all the wedges in Fig.8 have the same wedge angle, while the wedges in Fig.10 have different wedge angle. Wedges P1 and P2 in the first stage have a wedge angle of φ1, while wedges P3 and P4 in the second stage have a different wedge angle φ2 .
In thestructure #2 shown in Fig.8, wedges P2(P1) and P3(P4) constitute a parallel plate for two backward rays with SOP transformation o→e→e→o and e→o→o→e. Now in Fig.10, the wedge angles of P2(P1) and P3(P4) are different. They can’t constitute a parallel plate any more. Thus these two rays are also blocked.
Tracing of the two backward rays is shown in Fig.11. The circumstances in structure #2 and #3 are both shown for comparison. Each stage of the isolator core acts as a Wollaston prism for the backward light. In Fig.11(a) for structure #2, the second stage first separated the two rays angularly. Then the first stage turns the two rays back to the horizontal direction because the turning angles ϕ1 and ϕ2 by the two stages are the same. In Fig.11(b) for structure #3, the two rays are also first angularly displaced by the second stage and then turned by the first stage. However, the rays can’t be back to the horizontal direction because of difference on the angles ϕ1 and ϕ2, which results from difference on the wedge angles φ1 and φ2. As we can see in Fig.6, the isolation is rather sensitive to angular displacement. A small difference between φ1 and φ2 can generate enough Δϕ=ϕ2-ϕ1 to ensure the isolation. Note that there is another factor sin2δ (as shown in Fig.9) helping to block the two rays with SOP transformation of o→e→e→o and e→o→o→e.
Assembly of the dual-stage optical isolator core is shown in Fig.12. The wedges and FRs are sandwiched by two quartz substrates with arcuate intersection. Then all the components are housed in a magnet ring and fixed by adhesive.
Jing-Jong Pan, Ming Shih, and Jingyu Xu. Dual Stage Optical Device with Low Polarization Mode Dispersion and Wavelength Insensitivity. USA, US005581640A, 1996. 1~12
Wan Zhujun, Cao Mingcui, Ji Hangfeng, Isolation Analysis of In-Line Dual Stage Optical Isolator, Chinese Journal of Lasers, 29(11): 995~999, 2002
Written by Zhujun Wan, HYC Co., Ltd