Fully crystalline channel waveguide lasers are very promising as efficient and compact sources for many laser applications requiring significant power scaling with nearly diffraction-limited beam quality. In a double-clad implementation, they are direct analogs of conventional (glass-based) fiber lasers with the same ability to maintain high clad-pump intensity and tight confinement of both pump and laser modes along the entire length of the gain medium. Analysis of conventional glass fiber laser power scaling [1,2] indicates that, due to low glass thermal conductivity, heat generation associated with pumping is still strong. As a result, fibers may reach fracture threshold before nonlinear scaling limits, while nonlinearities, e.g., stimulated Brillouin scattering (SBS), will be prime power scaling limiters for very narrowband [especially single-longitudinal-mode (SLM)] fiber lasers. As opposed to conventional fibers, fully crystalline double-clad waveguides are more suitable for laser power scaling due to their $\sim 10$ times higher thermal conductivity as well as an order of magnitude higher absorption and emission cross sections of common rare-earth dopants [3]. In addition, yttrium–aluminum–garnet (YAG) waveguides have an extremely low SBS gain coefficient from $9\times {10}^{-15}$ to $5\times {10}^{-12}\mathrm{m}/\mathrm{W}$ (variation is based on data from different sources [3]), versus that of silica glass ($5\times {10}^{-11}\mathrm{m}/\mathrm{W}$). Estimates indicate that fully crystalline double-clad fiber lasers can be scaled to tens of kilowatts of power even in the most demanding SLM laser design [3,4]. Therefore, small-core waveguide laser development is of great importance.

Recently, a resonantly cladding-pumped, Er:YAG double-clad waveguide laser, fabricated by an adhesive-free bonding (AFB), has been demonstrated [5]. Taking advantage of the large numerical aperture (NA) of the inner cladding, the laser was efficiently pumped by a conventional laser diode bar stack with poor beam quality and delivered 25.4 W of continuous-wave (CW) output at 1645 nm with 56.6% optical efficiency (limited by high cladding loss). This double-clad crystalline laser had a free-space open cavity, and, due to the large core size ($500\mathrm{\mu m}\times 500\mathrm{\mu m}$), it was sufficiently multimode. A nearly diffraction-limited output has been demonstrated from a side-pumped multimode small-core ($100\mathrm{\mu m}\times 80\mathrm{\mu m}$) Yb:YAG channel waveguide [6]. This effort also used a free-space open cavity, and despite a very short (10 mm) waveguide length, the achieved slope efficiency was relatively low for resonant pumping (43%) due to imperfect matching of the pump and laser modes. To the best of our knowledge, the highest reported slope efficiency of 80.4% was achieved in a 22 μm thick $\mathrm{KY}{({\mathrm{WO}}_4)}_2\text{:}\mathrm{Yb}$ planar waveguide [7].

In this Letter, we demonstrate a small-core, resonantly pumped, single-mode, Er:YAG channel waveguide laser with diffraction-limited output and nearly quantum defect limited efficiency. This first core-pumped (${\mathrm{Er}}^{3+}\text{:}\mathrm{YAG}$-core/undoped YAG cladding) waveguide laser delivered a 9.1 W CW output with a slope efficiency of 92.8% (versus the absorbed pump power) and $M^2=1.05$.

The investigated 41.3 mm long channel waveguide, manufactured by AFB [8,9], had a $61.2\mathrm{\mu m}\times 61.6\mathrm{\mu m}$ cross section ${\mathrm{Er}}^{3+}(0.25\%)\text{:}\mathrm{YAG}$ core embedded in a $3\mathrm{mm}\times 5\mathrm{mm}$ YAG cladding; see Fig 1(a). The channel was buried in the cladding only 30 μm deep under the $5\mathrm{mm}\times 41.3\mathrm{mm}$ surface. The small refractive index difference, $\mathrm{\Delta }n$, between the Er-doped YAG core and undoped YAG cladding provides sufficient waveguiding with low NA [9]. This $\mathrm{\Delta }n$ was measured (with the accuracy of $\sim {10}^{-6}$) by the interferometric method with specially prepared adhesive-free bonded (Er:YAG/YAG) test samples for several ${\mathrm{Er}}^{3+}$ concentration values [10]. Given the refractive index of the undoped YAG cladding at 1532 nm of 1.8073 [11] and the measured $\mathrm{\Delta }n=5.2\times {10}^{-5}$ between the ${\mathrm{Er}}^{3+}(0.25\%)\text{:}\mathrm{YAG}$ and undoped YAG, the core NA was derived to be ${\mathrm{NA}}_C\sim 0.014$. Such an ultralow NA should lead to a single transverse mode operation of the channel waveguide despite its relatively large (on the scale of step-index conventional fibers) core cross section. However, the ultralow NA presents a challenge for achieving efficient pump launching into the core from sources with large angular divergence. To mitigate this problem, a single mode Er-fiber laser was used as a pump source. Both ends of the Er:YAG waveguide were antireflection (AR) coated for both laser and pump wavelengths ($R=0.25\%$ at 1525–1650 nm). The waveguide was mounted on a water-cooled plate and conductively cooled at 18°C.

 

Fig. 1. (a) Er:YAG/YAG waveguide rendering (not to scale). (b) Experimental layout of the Er:YAG-core waveguide laser pumped by a single-mode Er-fiber laser. (c) Optical layout used for simultaneous measurements of the incident and transmitted pump power along with the laser output power.

Download Full Size | PDF

The experimental layout is shown in Fig. 1(b). A CW, 20 W rated Er-fiber laser, with $\sim 0.3\mathrm{nm}$ full width at half-maximum (FWHM) at $\sim 1532\mathrm{nm}$, was used to longitudinally pump the Er:YAG core. A collimated, unpolarized pump with a divergence of $\sim 1\mathrm{mrad}$ was focused into the waveguide by an aspheric lens $L_1(f_1=50\mathrm{mm})$ through a plano–plano dichroic mirror (HR at 1600–1650 nm, AR for the pump). The lens provided a $\sim 50\mathrm{\mu m}$ pump spot diameter that fits well into the waveguide core cross section of $61.2\mathrm{\mu m}\times 61.6\mathrm{\mu m}$. This pumping geometry corresponded to an ${\mathrm{NA}}_F$ $\sim 0.019$ of the incident pump beam and yielded the best pump launching efficiency.

The laser cavity was formed by the above-mentioned dichroic mirror and a plano-concave output coupler (OC). Both mirrors were placed as close as possible ($\sim 40\mathrm{\mu m}$) to the waveguide ends. This configuration provided the maximum output power. An analysis of open-cavity modes shows that excitation of the ${\mathrm{TEM}}_{00}$ laser mode in such a cavity is impossible due to the high diffraction loss, and only a waveguiding mode can be supported [12].

The laser beam divergence was determined with an IR CCD camera (Spiricon, model LW230) placed in the focal plane of the spherical lens with a focal length of 150 mm.

Due to the short length of the waveguide, not all of the incident power is absorbed on the first pass. In order to derive the maximum achievable optical-to-optical efficiency, one has to measure the unabsorbed pump power reflected off the OC back into the waveguide. Figure 1(c) depicts the optical layout used for simultaneous measurements of the incident and transmitted pump power along with the laser output power. The accuracy of the entire derivation also relies on precise knowledge of the transmissions and reflections of all optical components used in the setup shown in Fig. 1(c).

In the case of single-pass pumping schemes, we determined the absorbed pump power ($P_{\text{abs}1}$) by measuring the incident pump power ($P_0$) and the residual pump power behind the OC ($P_{\text{res}}$) when the laser is above the threshold; see Fig. 1(c):

The first pass pump absorption $A_1$ is determined as

This same method cannot be easily used for the second pass (in the counterpump direction) due to the poor measurement accuracy. Instead, we assumed that the relative pump absorption in the waveguide, defined as the ratio of the absorbed pump power to the incident one, $\alpha =P_{\text{abs}}/P_{\text{inc}}$, was equal for both beam directions; i.e., $\alpha =A_1$. Such an assumption is acceptable when the pump saturation effect is negligible, which was the case in our experiments. This approach can only overestimate the total absorbed pump power because pump absorption on the second pass can only be lower due to the counterpropagating beam geometry, but not higher. The close proximity of the OC to the waveguide end simply ensures that the pump beam is reflected right back into the waveguide core.

The pump incident on the core entrance end further propagates in two distinct ways: (i) the pump that becomes confined inside the core (launched pump) is absorbed over the entire core length before leaving through the opposite end; (ii) the pump that escapes the core due to the mismatch between the pump NA (${\mathrm{NA}}_F$) and the core NA (${\mathrm{NA}}_C$) (where ${\mathrm{NA}}_F\sim 0.019$, ${\mathrm{NA}}_C=0.014$) propagates in the cladding. The launching efficiency, to a first approximation, scales as the squared ratio of the core and the pump beam NAs, and in our case is ${({\mathrm{NA}}_C/{\mathrm{NA}}_F)}^2=0.54$.

Figure 2(a) shows the CW output power of the Er:YAG waveguide laser versus the incident pump power at 1532 nm. The maximum output power of 9.1 W and the best optical-to-optical efficiency of 49.2% were achieved with $R_{\mathrm{OC}}({\lambda }_{\text{LASER}})=80\%$, $R_{\mathrm{OC}}({\lambda }_{\text{PUMP}})=85\%$, and $\mathrm{RoC}=50\mathrm{cm}$. With this OC full absorption of the launched power was achieved by double passing the 41 mm long Er:YAG core. On this condition, Fig. 2(b) depicts the output power of the waveguide laser plotted versus the absorbed pump power. The observed laser efficiency is near the quantum defect limited value for resonant pumping at 1532 nm.

 

Fig. 2. (a) CW output power of the resonantly pumped Er:YAG-core waveguide laser versus the incident pump power. (b) CW output power versus the absorbed pump power. Laser cavity: length $Z=41.3\mathrm{mm}$, $R_{\mathrm{OC}}({\lambda }_{\text{LASER}})=80\%$, $R_{\mathrm{OC}}({\lambda }_{\text{PUMP}})=85\%$, and $\mathrm{RoC}=50\mathrm{cm}$.

Download Full Size | PDF

The Er:YAG-core waveguide was designed to be a single-mode device. The criterion of single-mode operation for square waveguides is $B<1.37$, where the $B$ number is the “square analog” of the $V$ number for round fibers [13]:

The measured beam diameter as a function of the position of the imaging plane relative to the location of the beam waist is plotted in Fig. 3. It can be seen that the collected data are in agreement with predictions based on Gaussian beam divergence with $d_0$ set to 67 μm:

The beam quality $M^2$ derived from the above measurements was around 1.05, which proves that our laser is a purely single-mode device.

 

Fig. 3. Er:YAG waveguide laser beam diameter at $e^{-2}$ level as a function of the imaging plane position near the focal plane of the lens $f=150\mathrm{mm}$.

Download Full Size | PDF

The ${\mathrm{Er}}^{3+}\text{:}\mathrm{YAG}$ waveguide laser operated simultaneously at two wavelengths, 1617 and 1645 nm, if the $R_{\mathrm{OC}}$ was greater than 75%. It operated at 1617 nm only if the $R_{\mathrm{OC}}$ was below 70%. This observation is in agreement with the well-known interplay between the room-temperature emission cross sections and ground-state absorptions of the 1617 and 1645 nm transitions and population inversion of ${\mathrm{Er}}^{3+}$ in YAG (best explained in [14]). The maximum bandwidth of the laser emission at the maximum pump power was $\sim 3.6\mathrm{nm}$ (FWHM) for both 1617 and 1645 nm laser lines.

The laser output was linearly polarized parallel to the 5 mm side of the cladding facet. The polarization extinction ratio was around $1000:1$. This result can be tentatively explained by the asymmetric positioning of the buried channel inside the cladding described above. A corresponding asymmetric cooling causes the observed polarization preference.

In summary, we demonstrated the CW operation of a resonantly pumped (${\mathrm{Er}}^{3+}\text{:}\mathrm{YAG}$-core/undoped YAG cladding), single-mode channel waveguide laser with nearly quantum defect limited efficiency. A core-pumped waveguide delivered 9.1 W of output power at 1617 and 1645 nm with a slope efficiency of 92.8% (versus the absorbed pump power) and a nearly Gaussian output beam with $M^2=1.05$. The demonstrated efficiency is believed to be the highest optical-to-optical efficiency ever demonstrated for a channel waveguide laser.