Crystal-based Devices — High-power Optical Isolators

The research and application of fiber lasers flourish in recent years. Optical isolators are important devices to ensure the reliability of fiber lasers. Depending on the actual applications, different types of optical isolator are employed. For some applications under relatively low optical power, wedge-type in-line optical isolator is employed, just with more consideration on heat dissipation. While for some applications under high optical power, BD-type in-line optical isolator is employed instead of wedge-type. The reason is that the FR for telecom applications can’t be used anymore.

Why Is BD-type Selected?
The FR for telecom applications is a thin film of Bi-substituted rare-earth iron garnet single crystals (RIG) grown by liquid phase epitaxy. The RIG film will be damaged by high power laser. Thus TGG (terbium gallium garnet) crystal is used for optical rotation as the substitution due to its endurance to high optical power. The rotatory angle by the TGG crystal depends on Verdet constant, crystal thickness and magnetic field strength, as Eq. (1). Usually, the magnet field is not uniform, thus Eq. (1) is modified as Eq. (2).

For the RIG FR in telecom applications, the rotatory angle is fixed under a saturated magnet field. The actual rotatory angle depends on the thickness of the RIG film epitaxially grown on a substrate. The film thickness is doubled for a 90º rotator than a 45º rotator. The thickness of a 45º FR @1550nm is about 0.36mm (including the substrate thickness). A RIG FR just needs a small magnet ring providing magnetic field more than saturation (usually 200Oe).

While for a TGG rotator, the magnet ring needs to be elaborately designed to provide a precise rotatory angle. Moreover, the thickness of TGG crystal (10~15mm) is much more than that of RIG FR. If the TGG rotator is used in the wedge-type isolator, the lateral offset and walk-off of the rays are too much. The assembly of the device is difficult and the two forward rays can’t be received by the output collimator simultaneously.

Thus we know, for high power applications, TGG is required instead of RIG. The thickness of 10~15mm excludes wedge-type structure. BD-type is selected as substitution.

Design of TGG Rotator
The Verdet constant of TGG crystal is V=0.002º/(G·cm). According to Eq. (2), a high strength magnetic field is required, which is to be generated by a big magnet. The TGG rotator (including magnet rings and TGG crystal) need elaborate design to obtain the required strength of magnetic field with a minimal magnet.

The magnet ring for a RIG FR is usually axially magnetized, as shown in Fig.1(a). It generates a magnetic field as show in Fig.1(b). The magnetic field strength at the central axis is given by Eq. (3) and is simulated as curve (a) in Fig.4. The RIG FR is usually placed at the central of the magnet ring, where the magnetic field is nearly uniform and saturated.

(3)

Where Br is the residual magnetic field of the material, 2La is the length of the magnet ring, Ra1 and Ra2 are the inner and outer radii, respectively.

Fig.1 Axially magnetized magnet ring and the magnetic field it generates
Fig.1 Axially magnetized magnet ring and the magnetic field it generates
Fig.2 Radially (divergent) magnetized magnet ring and the magnetic field it generates
Fig.2 Radially (divergent) magnetized magnet ring and the magnetic field it generates
Fig.3 Radially (convergent) magnetized magnet ring and the magnetic field it generates
Fig.3 Radially (convergent) magnetized magnet ring and the magnetic field it generates
Fig.4 Magnetic field strength along the central axis (generated by different magnets)
Fig.4 Magnetic field strength along the central axis (generated by different magnets)

However, a TGG rotator needs far more strength of magnetic field than a RIG FR. If only an axially magnetized magnet ring is used, the size will be too big for application. Radially magnetized magnet rings are introduced to generate magnetic field more efficiently. The radial magnetization of magnet rings is shown in Fig.2(a) and Fig.3(a). The former is magnetized in a divergent magnetic field, while the latter is in a convergent magnetic field. The magnetic field generated by a divergently and radially magnetized magnet ring is shown in Fig.2(b), while that generated by a convergently and radially magnetized magnet ring is shown in Fig.3(b). The distribution of the magnetic fields is identical with only difference in directions.

Along the central axis, the magnetic field strength of an axially magnetized magnet ring is given by Eq. (4) and is simulated as curve (b) and (c) in Fig.4.

(4)

Where Br is the residual magnetic field of the material, 2Lc is the length of the magnet ring, Rc1 and Rc2 are the inner and outer radii, respectively.

The central axis magnetic fields generated by divergently and convergently magnetized magnet rings are centrosymmetrical. The peaks of the magnetic field strength are at the edges of the magnet ring. The two peaks at the opposite edges are opposite in direction. As we can see in Fig.4, if the curve (c) is shifted to the right and the curve (d) is shifted to the left, they can be added to the curve (a). The superimposition of the magnetic field generated by an axially and two radially (one divergently and the other convergently) magnetized magnet rings is given by Eq. (5) and simulated as the curve (d) in Fig. (5). For comparison, the magnetic fields generated by the axially and radially (shifted left and right) magnetized magnets are also given as the curves (a-b).

(5)

Fig.5 Magnetic field strength along the central axis (generated by magnet assembly)
Fig.5 Magnetic field strength along the central axis (generated by magnet assembly)

As we can see in Fig.5, the assembly of three magnets generates stronger magnetic field at a broadened space. When the TGG crystal with length L is placed at the center of the magnet assembly, the rotation angle is given by Eq. (6).

The rotation angle θ needs to be 45º exactly. Changing the length of the radially magnetized magnets and the TGG crystal an effective approach to make minor adjustment on θ. The precise parameters can be experimentally obtained based on simulation.

Some Engineering Considerations
1)Processing of Fiber Endface

In fiber laser applications, the input of an optical isolator is usually a few mode fiber (FMF) with core diameter ~20μm. The mode field diameter (MFD) of the fundamental mode is ~16μm. The input optical power is usually tens of Watt, which is too high and tends to damages the optical elements. Most damage happens at the fiber endface, where the optical power density is the highest. Any dust or defect at the fiber endface will absorb optical power and accumulate heat. Thus the power density should be reduced before emitting from the fiber endface, which is usually realized by expanding the emitting MFD from the fiber endface.

The first solution is to splice a quartz rod or multimode fiber (MMF) at the endface of the FMF, as shown in Fig.6. In the quartz rod of Fig.6 (a), the optical beam propagates with no confinement by any waveguide. Thus the MFD expands quickly. For a laser beam with MFD=16μm (@1.064μm) at the FMF endface, the MFD expands to 60μm after propagation distance of 1mm in the quartz rod. In the MMF core of Fig.6(b), the laser beam emits from the MMF endface before reaching the core/cladding interface. Thus the optical beam also propagates with no confinement by any waveguide, just as splicing with a quartz rod. The FMF and the spliced quartz rod (or MMF) are housed in a quartz capillary. When the MFD is expanded from 16μm to 60μm, the power density is reduced to 1/14, which help to avoid damage on the fiber endface.

Fig.6 Output MFD expanding by splicing of quartz rod or MMF
Fig.6 Output MFD expanding by splicing of quartz rod or MMF

The second solution is to have the fiber endface physically contacted with a sapphire plate, as shown in Fig.7. The FMF is fixed in a ceramic ferrule and the endface is spherically polished, just like a fiber connector. The ferrule is pushed by a spring force and the spherical endface deforms slightly. Thus physical contact (PC) between the fiber endface and the sapphire plate is kept. The refractive indices (RI) of quartz and sapphire are 1.4496 and 1.7545, respectively. Thus 0.9% (20.45dB) reflection happens at the PC interface.

If another glass material with RI close to quartz is selected for substitution of the sapphire plate, the reflection can be reduced. However, the hardness and stability under pressure are not assured.

Fig.7 Output MFD expanding by physical contact with a sapphire plate
Fig.7 Output MFD expanding by physical contact with a sapphire plate

TEC fiber can also expand the MFD, while it is not applicable in high power optical isolator. Minor defects are inevitable during the TEC process. As we know, any defect will be the center of heat accumulation.

2)Setting of Apertures
I many applications of fiber lasers, the emitting laser beam is collimated, as shown in Fig.8. The collimator (usually a telescope composed of two lenses with short and long focal lengths) is mounted behind the optical isolator.

Fig.8 High-power optical isolator with collimated beam output
Fig.8 High-power optical isolator with collimated beam output

The light path in the optical isolator is shown in Fig.9. Two revised BDs are employed, which separate the o-ray and e-ray symmetrically and help to keep the input and output optical beam collinear. An aperture is usually placed at the entrance to block the backward light. Although the backward light deviates from the axis which makes it impossible to be received by the input fiber collimator, the input aperture helps to further improve the isolation.

Fig. 9 Light path in the high-power optical isolator
Fig. 9 Light path in the high-power optical isolator

Fiber lasers are widely used for laser marking, laser cutting and thermal treatment. A focusing lens and two rotating mirrors are usually employed. The surface to be marked, cut or thermally treated is placed at the focal plane of the lens, as shown in Fig.10. The two rotating mirrors are left out in the figure because they don’t affect the following analysis. The surface to be processed is usually not a mirror plane. The roughness means that many local surfaces with different normal directions exist. The focused beam is inevitably to scan on some special local surfaces, which will reflect the backward light to the entrance. Based on polarization analysis, we can see that p-wave and s-wave are reflected (by two different local surfaces) backward to the entrance, as shown in Fig.10(a) and (b), respectively. The backward light will reach the fiber laser and lead to failure of the equipment. During the long time working of fiber lasers, such special local surfaces are predestinate and thus failure is inevitable.

Fig.10 Failure mechanism in applications: reflection by local surfaces at the focusing plane
Fig.10 Failure mechanism in applications: reflection by local surfaces at the focusing plane

Based on above analysis of failure mechanism in applications, the solution is rather simple. One more aperture is placed at the output, as shown in Fig.12. The backward light reflected by the special local surfaces is blocked by the output aperture.

Fig.11 Solution for failure in applications: adding an output aperture
Fig.11 Solution for failure in applications: adding an output aperture

A perfect route of theoretical design, problems in applications, theoretical analysis, engineering solutions. Enjoy your success in engineering work!

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s