High-Brightness Ce 3+ :YAG Crystal Fiber Light Source

In Fig. 3.6, only the forward fluorescence was collected by L3. In order to fully utilize the fluorescence power and the pump power, the optical setup was changed to Fig. 3.9.

The pump light was reflected by a dichroic beam splitter (LM01-466, Semrock), DBS, and focused into the crystal fiber. The collimation lens and the focusing lens are identical to the ones used in the previous section. A broadband dielectric mirror (CVI BBD1-PM-1037-C), BBM, was attached on the opposite end of the crystal fiber to reflect both the pump light and the forward fluorescence light. The reflectance of the broadband mirror is

>99% from 488 to 694 nm and about 95% at 450 nm. Dielectric coated mirror was chosen instead of metal mirror due to the higher damage threshold. The reflected forward fluorescence, together with the backward fluorescence, was collimated by L2 and transmitted through the dichroic beam splitter. Ideally, assuming perfect alignment and no propagation loss, attaching the BBM can increase 20% pump absorption and increase 100% fluorescence by reflecting the forward fluorescence. The expected overall increase would be (1+20%) × (1+100%) - 100% = 140%. For comparison between the forward and the backward fluorescence power, the forward fluorescence was also measured by removing the BBM and used the same measurement setup after the crystal fiber in Fig.

3.6. The crystal fiber used in this experiment was the one measured to have the highest conversion efficiency in the previous experiment. The fiber was 2-cm long and had a core diameter of 25 μm.

The measured fluorescence powers are shown in Fig. 3.10. The powers under 1.4-W incident pump power are 14.2 m1.4-W and 13 m1.4-W for the backward fluorescence without BBM and the forward fluorescence, respectively. The difference can be explained by smaller propagation loss experienced by the backward fluorescence than the forward one.

Because the most of the fluorescence was emitted near the pumping end where the pump intensity is higher than the opposite end, the averaged propagation distance before exiting the crystal fiber would be shorter for the backward fluorescence. The conversion efficiency drops with the increasing pump power, from about 2% under 100-mW pump to 1.45% under 1.4-W pump. The reason of this efficiency decay is likely to be the thermal quenching [29].

After attaching the BBM, the highest output power under 1.4-W incident pump power came to 19.9 mW. The enhancement with respect to the backward case without the BBM is 40%, much lower than the ideal value of 140%. The results indicates that the contact between the BBM and the crystal fiber endface was not good. Inspection of the crystal fiber end face under the optical microscope revealed that, although the packaged crystal fiber had been polished to optical quality, the surrounding tin surface is still not flat within the 1 cm × 1 cm area. This problem can be overcome by using a bare crystal fiber, but the thermal dissipation may not be as good as the tin package even if thermal grease were used.

450-nm LD L

PD DBS

L1

Ce³⁺:YAG crystal fiber BBM

L2

Fig. 3.9. The backward fluorescence measurement setup. BBM: broadband mirror. L1 and L2: aspheric lenses. DBS: dichroic beam splitter. LD: laser diode. PD: photo diode. Solid arrow: pump propagation direction. Hollow arrow: fluorescence propagation direction.

0 500 1000 1500

Fig. 3.10. The output fluorescence power versus the incident pump power.

The backward fluorescence spectra with and without the BBM are measured with a miniature spectrometer (USB4000-UV-VIS, Ocean Optics), as shown in Fig. 3.11. The normalized spectra are very similar because of the flat reflectance spectrum of the BBM in the fluorescence wavelength range. The 3-dB bandwidth of both spectra are 105 nm.

500 550 600 650 700

Fig. 3.11. Normalized fluorescence spectra of the Ce³⁺:YAG backward fluorescence with and without BBM.

During the experiment it was found that the collimation lens, L2, has severe dispersion, thus the collimation of the output fluorescence was not good. Using an objective lens can correct this problem, at the expense of higher cost and lower

transmittance.

By assuming 98% transmittances for the collimation lens L2, the radiance at the crystal fiber output was calculated to be 21.57 mW sr^-1. The calculated brightness (radiance) distribution is shown in Fig. 3.12. The maximum radiance is 32.6 W mm^-2 sr^-1 at 53.3 degree away from the axial direction, and 30.3 W mm^-2 sr^-1 in the axial direction.

Fig. 3.12. Radiance distribution of the high-brightness crystal fiber light source. The unit of the radial scale is W mm^-2 sr^-1.

The output powers and radiances of the Ce³⁺:YAG crystal fiber light source and several high-power LED light sources from Thorlabs are listed in Table 3.1. The viewing angle was defined as the full-width-half-maximum of the radiance intensity distribution.

Although the power of the LEDs are high, the crystal fiber light source is still 53 times brighter than the brightest LED due to the much smaller emitting size of the crystal fiber.

Table 3.1. Characteristics of Ce:YAG crystal fiber light source and Thorlabs high-power LED light sources. [56]

Light source λp

a Collimated output power measured with a 0.65-NA collimation lens

b Total output power measured with an integrating sphere

10

The high brightness light source would be very useful if it can be coupled into communicational glass fibers, taking advantages of the wide variety of glass fiber technologies. An experiment was conducted to couple the light source into a corning SMF28e fiber. The setup is illustrated in Fig. 3.13. An aspheric lens, L3, with f = 2.8 mm and NA = 0.65 (5721-A-H, New Focus) was used for coupling into SMF28e.

450-nm LD L

DBS

Ce³⁺:YAG

crystal fiber L2

SMF28e

L1

L3 PD

BBM

Fig. 3.13. Setup of SMF28e-pigtailed Ce³⁺:YAG light source.

The fluorescence power output from SMF28e is shown in Fig. 3.14. The maximum output power is 47.1 μW. The fluorescence spectrum is shown in Fig. 3.15. The 3-dB bandwidth is 96 nm, slightly narrower than the input spectrum. The dispersion of the aspheric lenses L2 and L3 is likely to be the cause of the spectral narrowing.

0 500 1000 1500

0 10 20 30 40 50

Fluorescence (µW)

Pump power (mW)

Fig. 3.14. Fluorescence power output from SMF28e fiber.

500 550 600 650 700 0.0

0.2 0.4 0.6 0.8 1.0

Normalized spectral density (a. u.)

Wavelength (nm) 3-dB BW: 96 nm

Fig. 3.15. Fluorescence spectrum output from SMF28e fiber.

Chapter 4

Broadly Tunable and Low-Threshold Cr

:YAG Double-Clad Crystal Fiber Lasers

In the previous work of [23], efficient Cr⁴⁺:YAG crystal fiber laser has been demonstrated with an 78.2-mW threshold pump power with only passive cooling . The laser cavity was formed by multilayer coatings deposited on both ends of the crystal fiber and the output wavelength was free-running at 1420 nm. For developing a tunable laser, the cavity must be extended to outside of the crystal fiber, so a wavelength selective element can be incorporated.

In this chapter, we present the first demonstration of external-cavity double-clad crystal fiber lasers with low pump thresholds and high slope efficiencies. The experimental results of broadly tunable crystal fiber lasers were also presented. In the last section, the performance of an optical amplifier based on the Cr⁴⁺:YAG double-clad crystal fiber was simulated.

4.1 Fabrication of Cr

:YAG Double-Clad Crystal Fibers for Laser Applications

4.1.1 Double-Clad Crystal Fiber Growth

The single-crystal fibers were fabricated by the LHPG method with steps similar to those introduced in 3.1.2. A Cr⁴⁺:YAG single-crystalline rod with 500 μm × 500 μm cross section was grown to a single crystal fiber with 68-μm diameter in a two-step diameter-reduction process by the LHPG system. The 68-μm single crystal fiber was inserted into a fused-silica capillary with a 76-μm inner diameter and a 323-μm outer diameter. The fiber-filled capillary was sealed by the CO2 laser of the LHPG system at one of its ends, and vacuumed through the other end for 30 min. Since the gap between the crystal fiber and the inner wall of the capillary is as small as several micrometers, to facilitate the removal of the air in the gap, the fiber-filled capillary was pre-heated by using a CO2 laser at a mild power to scan along the capillary. The residual air in the gap was heated and escaped out of the capillary. With the pre-heat process, the yield of the double-clad fiber growth process was significantly improved since the bubbles in the fiber caused by the

residual air were avoided.

The filled capillary was then fused by the sapphire-tube-assisted co-drawing LHPG (CD-LHPG) process [21]. Before we introduce the sapphire-tube assisted CD-LHPG process, it is worth to describe first the CD-LHPG process without sapphire tube. In this process, the pre-heated capillary was heated by the CO2 laser and attached to the crystal core for cladding formation. Since the 1600 °C softening point of the fused silica is close to the 1960 °C melting point of the Cr⁴⁺:YAG crystal, both central part of the fused silica capillary and the outer part of Cr⁴⁺:YAG crystal fiber melted during the growth process.

The molten fused silica mixed with the molten crystal and formed an additional inner cladding layer in between the crystal core and the cladding, as shown in Fig. 4.1. Because part of the crystal core was consumed by the formation of the inner cladding layer, the core diameter was reduced.

Fig. 4.1. An illustration of the CD-LHPG growth of the Cr⁴⁺:YAG double-clad crystal fiber. The inset photo showing the funnel-shpaed molten zone was captured during the fiber growth.

The sapphire-tube-assisted CD-LHPG technique was developed for enhancing the core diameter uniformity and reducing the propagation loss. In the CD-LHPG process, the uniformity of the core diameter was sensitive to the power stability of the CO2 laser.

To reduce this sensitivity, a sapphire tube with a 500-μm inner diameter of and a

1200-μm outer diameter was used to serve as a heat capacitor. The sapphire tube was placed at the focal spot of the CO2 laser, and the fused silica capillary filled with a single crystal fiber went through the center of the sapphire tube. The sapphire tube was heated by the CO2 laser and generated strong thermal radiation to melt the filled capillary. The stability of the thermal radiation from the sapphire tube is better than that of the CO2 laser, due to the thermal capacity provided by the sapphire tube. Therefore, this indirectly heating approach can produce crystal fibers with better core uniformity and lower propagation loss. Besides, compared with the focal spot of the CO2 laser, the heated sapphire tube provides longer heating time, which could benefit the crystal quality due to the annealing effect. The smallest core diameter that can be stably produced with the sapphire-tube-assisted CD-LHPG process is 10 μm, half of the smallest single-crystal fiber diameter can be achieved with the conventional LHPG process. The measured propagation loss of the double-clad crystal fiber with 10-μm core was 0.02 dB/cm [21].

Fig. 4.2. An illustration of sapphire-tube-assisted LHPG system.

A typical index profile of the Cr⁴⁺:YAG double-clad crystal fiber is shown in Fig.

4.3. The index profile was measured by a laser scanning microscope with a 635-nm laser source [57]. The refractive indices of the core and the outer clad are 1.82 and 1.46, respectively. The refractive index of the inner clad changes gradually from 1.60 near the outer clad to 1.66 near the core.

-80 -60 -40 -20 0 20 40 60 80 1.4

1.5 1.6 1.7 1.8

Refractive index (a. u.)

Position (µm)

Fig. 4.3. Refractive index profile of a Cr⁴⁺:YAG double-clad crystal fiber.

The optimal core diameter for minimizing the residual strain inside the double-clad crystal fiber is about 20 μm [58]. Although a smaller core diameter can effectively reduce the pump threshold, a large deviation from the 20-μm core diameter can significantly reduce the laser efficiency due to the large residual strain. In this dissertation, double-clad crystal fibers with core diameters from 17 to 19 μm were used for laser experiments.

4.1.2 Dielectric Coating Deposition

The as-grown double-clad crystal fiber was examined under the optical microscope.

Regions with defects or core diameters which did not meet the requirements were marked with the marker pen. A section of crystal fiber with desired length was then cut out with a diamond-tip scribe, and the marked regions were avoided. In the next step, the crystal fiber was fixed in a specially designed aluminum holder with a thermoplastic adhesive (Crystalbond 509). Finally, both surfaces of the packaged sample were ground and polished for dielectric coating deposition.

In order to build up an external-cavity crystal fiber laser, one end of the crystal fiber was deposited with highly reflective (HR) coating and the other end with anti-reflective (AR) coating. The HR coating serves as the cavity mirror, and the AR coating reduces the intra-cavity loss and suppresses the sub-cavity formation. The coatings are SiO2/TiO2

stacks deposited by an E-gun deposition system at a substrate temperature of 280 °C. The spectra of the coating on the crystal fiber endface was derived from the measured transmission spectra of the test glass plates, as shown in Fig. 4.4. The transmission of the

HR coating at the 1064 nm pump wavelength was as high as 94.8%, and the transmission in the 1350 to 1500 nm laser operation window was below 0.6%. The AR coating has a transmittance of 97.4% at 1064 nm and above 99.3% in the 1350 to 1500 nm wavelength range. The microscope photo of the coated fiber endface is shown in Fig. 4.5. The core and inner cladding regions were free from defects.

10000 1200 1400 1600

20 40 60 80 100

Transmittance (%)

Wavelength (nm) Output end Input end

Fig. 4.4. Transmission spectra of the coatings on both ends of the crystal fiber.

(a) (b)

Fig. 4.5. End photo of coated crystal fiber.

(a) Input end (HR), (b) Output end (AR).

4.2 Fluorescence Lifetime Thermal Loading and Polarization-Dependent Gain of Cr

:YAG Double-Clad Crystal Fiber

4.2.1 Fluorescence Lifetime Thermal Loading

As previously discussed in section 2.2.1, The fluorescence lifetime of Cr⁴⁺:YAG decreases with the rising crystal temperature due to the increased non-radiative relaxation rate. With smaller fluorescence lifetime, higher pump power is required for achieving the same population inversion level and thus the laser efficiency reduces. The relation between the fluorescence lifetime, τ_f, is linearly dependent on the temperature in a limited temperature range between 0 to 100 °C, as shown in Eq. (2.17).

To investigate the thermal dissipation ability of the crystal fiber, the fluorescence lifetimes of the crystal fiber and the bulk crystal were measured. A mechanical chopper (MC1000A, Thorlabs) was used to modulate the pump beam at a 1-kHz rate. The 10-slot blade (MC1F10, Thorlabs) was placed at the focal spot to achieve a fall time of 375 ns.

Nine of the ten slots were blocked with aluminum foil to avoid chopping the pump beam at different axial position due to the blade deformation. The duty cycle of the chopped pump beam was 5%. The residual pump light was blocked by an edge filter. The fluorescence was detected by a 150-MHz InGaAs photodiode (PDA10CF, Thorlabs) and recorded by a digital oscilloscope (TDS320, Tektronix) with a sampling rate of 500 million samples per second.

Fig. 4.6. Fluorescence lifetime measurement setup. LD: laser diode; L1~L5: aspheric lenses; C: chopper; HWP: half-wave plate; LPF: long-wavelength-pass filter; PD: photo diode.

A 4.4-cm long <0 0 1> crystal fiber with a 19-μm core diameter was used in the experiment. The sample was mounted on a copper heat sink with silver gel for passive cooling. The measured fluorescence decay curve under the 400-mW incident pump power

is shown in Fig. 4.7. The fitted fluorescence lifetime was 4.17 μs. The fluorescence lifetime of a bulk Cr⁴⁺:YAG crystal with a size of 0.5 mm × 0.5 mm × 2 mm was also measured with the same setup. The dependence of the fluorescence lifetime on the absorbed pump power of both crystal fiber and the bulk crystal was shown in Fig. 4.8.

The data were measured with the incident pump powers of 10, 100, 200, and 400 mW. It is obvious that the fluorescence lifetime decreases slower with increasing pump power for the crystal fiber. The relative lifetime decreases from 10-mW to 400-mW incident pump were 1.8% and 9.8% for the crystal fiber and the bulk crystal, respectively.

Substituting these values into Eq. (2.17) and use a τ^T_f of 0.0416 μs/°C [40], the corresponding temperature rise was found to be 1.7 °C and 9.1 °C for the crystal fiber and the bulk crystal, respectively. This is a clear evidence of better thermal dissipation for the crystal fiber.

-5 0 5 10 15 20 25 30 35

10^-4 10^-3 10^-2 10^-1 10⁰

Normalized fluorescence

Time (µs) Measured

Fitted

Fig. 4.7. Decay of the fluorescence intensity after blocking-off the pump.

Blue dots: measured data. Red line: the exponential fit with a time constant of 4.17 μs.

0 40 80 120 160 200 240 3.9

4.0 4.1 4.2 4.3

Lifetime (µs)

Absorbed peak pump power (mW) Bulk

Crystal fiber 0 Absorbed average pump power (mW)2 4 6 8 10 12

Fig. 4.8. Fluorescence lifetime vs. absorbed pump power for bulk and crystal fiber.

4.2.2 Polarization-Dependent Gain

The polarization properties of the Cr⁴⁺:YAG crystal has been discussed in section 2.2.1. If the pump is polarized along one of the crystallographic axes, one of the three types of Cr⁴⁺ ions will be selectively excited, and the gain will be maximum for the signal light polarized along the same crystallographic axis. However, in the crystal fiber, the pump and the signal might be depolarized due to the fiber birefringence, and the gain would decrease consequently. In order to investigate the polarization properties of our crystal fiber, the following experiment was conducted for measuring the polarization dependent gain.

The setup for measuring the polarization dependent double-pass gross gain is depicted in Fig. 4.9. The pump source was a PM980-pigtailed 1064-nm diode laser (LD-1064-BF-600, Innolume). A 1410-nm SMF28e-pigtailed DFB laser was used as signal light source. A 4.8-cm long <0 0 1> crystal fiber with a 19-μm core diameter is prepared for this measurement. One end of the crystal fiber was coated with a multilayer dielectric coating with anti-reflection for pump light and high-reflection for the signal light. The other end was coated with broadband AR coating for both signal and pump. The polarized pump was coupled into the crystal fiber from the dichroic-coated end with a 400-mW power. The attenuated signal was polarized and then coupled into the crystal fiber at the opposite end. The power of the small signal coupled to the crystal fiber is -29.6 dBm. The

polarization of the pump and the signal can be independently rotated by the half-wave plates. After one round trip in the crystal fiber, the returned signal is split by a polarization-insensitive beam splitter and collected into an optical spectrum analyzer (70950B, Hewlett-Packard). The gross gain is defined as the ratio of the signal power with pumping and that without pumping. By examining with an infrared card we found that both the pump and the signal exhibited smooth few-mode patterns after output from the crystal fiber.

Fig. 4.9. The polarization-dependent gain measurement setup. LD: laser diode; L1–L5: lenses; HWP: half-wave plate; BS: beam splitter; ND: neutral density filter; P: polarizer; LPF: long-wavelength pass filter; OSA: optical spectrum analyzer

The measured gross gain curves are shown in Fig. 4.10 and Fig. 4.11. The gain was maximum when the polarizations ofboth pump and signal are aligned to the same <0 0 1> crystal axis. When the polarizations of the pump and signal were aligned to different

<0 0 1> crystal axes, the gain was nearly zero. There was some difference between the peak gain values of the two orthogonal <0 0 1> crystallographic axes, which should be identical theoretically. This can be explained by the stress inside the crystal fiber which originated during the cooling stage of the co-drawing LHPG process. The emission cross section of Cr⁴⁺:YAG is dependent on the strain [58]. If the stress inside the crystal fiber

<0 0 1> crystal axes, the gain was nearly zero. There was some difference between the peak gain values of the two orthogonal <0 0 1> crystallographic axes, which should be identical theoretically. This can be explained by the stress inside the crystal fiber which originated during the cooling stage of the co-drawing LHPG process. The emission cross section of Cr⁴⁺:YAG is dependent on the strain [58]. If the stress inside the crystal fiber