We experimentally reveal the pump-induced loss in a Ti:sapphire laser crystal with 451-nm indium gallium nitride (InGaN) laser diode pumping and show that 478-nm pumping can reduce such loss. The influence of the pump-induced loss at 451-nm pumping is significant even for a crystal that exhibits higher effective figure-of-merit and excellent laser performance at 520-nm pumping. We demonstrate the power scaling of a Ti:sapphire laser by combining 478- and 520-nm InGaN laser diodes and obtain CW output power of 593 mW.
© 2017 Optical Society of America
Ti:sapphire crystals with broadband gain bandwidth and good heat tolerance are the most suitable femtosecond light source and have been widely used in research. In general, a frequency-doubled Nd:YAG, , optically pumped semiconductor laser (OPSL) or a Yb-doped-fiber laser is used as a pump source, unfortunately preventing the downsizing and low-pricing of Ti:sapphire lasers. Low-cost and compact femtosecond light sources are necessary for accelerating various applications with ultrafast coherent light.
In recent years, indium gallium nitride (InGaN) high-power blue laser diodes (LDs) have been developed that provide multiwatt output power around a wavelength of 450 nm. In 2009, the first experimental demonstration of a CW Ti:sapphire laser directly pumped by the 452-nm InGaN LD was reported by Roth et al. . The same group successively demonstrated mode-locking with 13-mW average power and 142-fs pulse duration using a saturable Bragg reflector (SBR) . Young et al. demonstrated a Kerr-lens mode-locked Ti:sapphire laser pumped by two 445-nm LDs, whose pumping power reached 2 W, and obtained a 50-fs pulse width with an average power of 105 mW and a 15-fs pulse width with 70 mW . Thus, the development of InGaN blue LDs enables the construction of a compact and cost-effective mode-locked Ti:sapphire laser. On the other hand, Roth et al. simultaneously reported the increase of intracavity loss with 450-nm pumping, called pump-induced loss .
In 2013, a high-power green (520-nm) InGaN LD, whose output power reached , was developed, and it became possible to employ that green LD for Ti:sapphire laser pumping since 520 nm is closer to the absorption peak of a Ti:sapphire crystal. In 2014, for the first time, to the best of our knowledge, our group achieved a mode-locked Ti:sapphire laser pumped by a 520-nm LD using a semiconductor saturable absorber mirror (SESAM) . In 2015, Gürel et al. demonstrated a Kerr-lens mode-locked Ti:sapphire laser without a SESAM with 520-nm LD pumping . They used two 520-nm LDs, which were driven at higher injection currents than the rated current at 1.5 A to achieve a 3-W total pump power incident onto a crystal, and obtained a 450-mW average output power at a repetition rate of 418 MHz and a pulse width of 58 fs. However, excess drive currents may damage InGaN LDs, even in pulse-mode operation. Since the present green LD’s power remains limited to 1 W, some beam-combining technologies are necessary for scaling pump powers and increasing the Ti:sapphire laser output power.
In this paper, we experimentally confirmed the existence of pump-induced loss in 450-nm LD pumping that was addressed by Roth et al. , and we discussed its mechanism. We also observed that pump-induced loss can be reduced at 478-nm LD pumping and demonstrated wavelength multiplexed pumping using 521- and 478-nm LDs to increase the incident pump power to .
2. MEASUREMENT OF PUMP-INDUCED LOSS
A. Observation of Pump-induced Loss at 451-nm Pumping
Roth et al. reported that since no reduction was observed in the fluorescence intensity, the output power degradation was caused by excess loss in the gain medium at 458-nm pumping. Then they confirmed a pump-induced loss through experimental evidence where a gradual decrease in Ti:sapphire laser output power took place when 532-nm frequency-doubled laser pumping was overlapped by 458-nm LD pumping . They reported also that the deteriorated output power recovered when 532-nm frequency-doubled laser pumping was maintained for a few minutes after the 458-nm pumping was terminated. On the other hand, Durfee et al. reported that no such pump-induced loss (photo-darkening) at simple 445-nm LD-only pumping . We repeated the same experiments as those by Roth et al. to examine the existence of pump-induced loss at 451-nm LD pumping.
First, we pumped a Ti:sapphire crystal by 520-nm LD, and added 451- or 478-nm LD pumping and observed variations of output power. Figure 1 shows the experimental setup of a Ti:sapphire laser directly pumped by InGaN LDs. We used a green InGaN LD (Nichia) at 520 nm whose maximum output was 1 W, an InGaN LD (Nichia) at 451 nm whose maximum output power was 3.5 W, and a 478-nm InGaN LD (Nichia) whose maximum output power was 1 W. The beam qualities of these LDs were measured to be , , and for 451, 478, and 520 nm, respectively. The Ti:sapphire crystal was 2.5-mm-long and cut at Brewster’s angle. The figure-of-merit (FOM), defined by residual absorption at 514 and 800 nm, exceeded 200 (based on a value measured spectroscopically by the manufacturer), and the nominal doping concentration was 0.25 wt. %.
Figure 2 shows plots of CW output laser power at 790 nm with output coupling of 2.5%. The laser cavity was optimized at the highest pumping power at an absorbed pump power of 2.1 W using a 520-nm LD and a 451-nm LD, and then the pump power was gradually decreased in the two different orders where either the 451- or the 520-nm LD power was decreased first. The laser slope efficiency at 451-nm pumping is ; that at 520-nm pumping is . This reduction in the slope efficiency is partially explained by the difference in the quantum defect at each pumping wavelength: 451/790 versus 520/790. However, the actual degradation at 451-nm pumping is much larger than that estimated by the quantum defect factors. Since the beam quality of the 3.5-W 451-nm LD is worse than that of the 1-W 520-nm LD, the mode matching efficiencies for each pumping were, respectively, estimated as 57% and 83% at 451- and 520-nm pumping. However, this lower mode matching efficiency still cannot explain the low slope efficiency at 451-nm pumping that we obtained in Fig. 2. When linearly extrapolating the output laser power at two 3.5-W 451-nm LDs for pumping and achieving the highest pump absorption power of 4 W with this Ti:sapphire laser crystal (Fig. 2), the estimated highest laser power can be only . This power is disappointing for constructing a low-cost, high-efficiency Ti:sapphire laser with blue InGaN LDs.
Next, we repeated the same experiment reported by Roth et al.  to examine whether such low efficiency at 451-nm LD pumping is caused by induced absorption. The temporal response of the laser output power with dual-wavelength pumping is shown in Fig. 3. Initially, only the 520-nm LDs pumped the Ti:sapphire laser with an absorbed power of 661 mW. The output power reached 18 mW, and no output degradation took place. By adding 451-nm LD pumping with absorbed power of 1460 mW, the output power initially increased to 70 mW but significantly decreased for and reached a steady-state power of . When the 451-nm pumping was blocked, the output power decreased less than the initial 18 mW and slowly recovered to 18 mW. This response of the laser output resembles that of Roth’s experiment where dual-wavelength pumping with 532 and 458 nm was used in the same temporal sequence .
In Fig. 4, we compared slow power degradations for 451-nm absorbed powers of 0.5, 1.0, 1.5, and 2.0 W. The output power exponentially decreased and reached a steady-state power. At after the 451-nm LD overlapping, was dropped by 0.12, 0.14, 0.15, and 0.18 for 451-nm pumping powers of 0.5, 1.0, 1.5, and 2.0 W, respectively. We also examined the temporal change of the output laser power at 451-nm LD pumping alone. No gradual power reduction occurred, as shown in Fig. 5.
Next, we repeated the same experiment as that in Fig. 4 with a 1-W 478-nm LD instead of the 451-nm LD. The temporal response of the laser output power with dual-wavelength pumping is shown in Fig. 6. We did not observe any degradation of output power when we added 478-nm pumping to 520-nm pumping. This result also agrees with the experimental results obtained with 476.5-nm pumping by Roth et al. .
B. Influence of Pump-induced Loss on Laser Performance for Different Laser Crystals
To quantitatively evaluate the effect of a pump-induced loss on the output laser performance, we measured the threshold pump power and laser slope efficiencies for pump wavelengths of 451, 478, and 520 nm with three different laser crystals and estimated the intracavity losses. The intracavity losses were determined by Findlay–Clay analysis  or the analysis scheme reported by Caird et al. . However, these analysis schemes can treat only a static residual intracavity loss, not a pump-induced loss. The amount of pump-induced loss depends on the excited state population, which is clamped at the population corresponding to the cavity threshold during lasing and on the pumping power. For the Findlay–Clay analysis, we estimated the threshold pump powers for various output couplers, whereas for Caird analysis, we measured the slope efficiencies for various output couplers. At higher output coupling (lower output mirror reflectivity ), the threshold power increases, leading to higher excited state populations. Therefore, the pump-induced loss may be slightly larger at higher output coupling.
We analyzed the laser performance with three laser crystals grown by the different manufacturers: Crystal A (GT Advanced Tech.), B (Castech Inc.), and C (Roditi Int.). The nominal doping concentration was 0.25 wt. % for all the crystals. The crystal length was 2.5 mm for Crystals A and B, while Crystal C was 4-mm long. We used a pumping optics as shown in Fig. 7, where two 520-nm LDs and two 478-nm LDs whose output power was 1 W were used. The pumping lasers were collimated by aspheric lenses and spread in the slow axis direction by a pair of cylindrical lenses and a Ti:sapphire crystal pumped from both sides. At 451-nm pumping, a single 3.5-W LD was used only from a single side of the cavity. The numerically calculated mode matching efficiencies for Crystals A and B were for 520- and 478-nm pumping, and for 451-nm pumping, respectively. For 4-mm-long Crystal C, the mode matching efficiencies were for 520- and 478-nm pumping, and for 451-nm pumping, respectively.
Figure 8 are plots of output laser power measured for various output couplers at the pump wavelengths of 520, 478, and 451 nm. The slope efficiencies and the threshold pump powers were summarized in Table 1. It is obvious that the laser slope efficiencies were degraded as the shorter pump wavelength was used for all three laser crystals. (We adjusted the scale of each figure in Fig. 8 so that one can directly compare the slope efficiencies by the tilt of the linear plots.). Although no transient power decrease was observed at the measurement shown in Fig. 6, laser power degradation existed even at 478-nm pumping compared to those at 520-nm pumping.
The highest fractions of the heat dissipation determined by the quantum yields without lasing is 34%, 40%, and 43% for 520-, 478-, and 451-nm pumping, respectively. However, we did not observe any thermal lens effects in those cavity conditions to achieve stable laser oscillation. Moreover, in Fig. 8, the output laser power always linearly increased as the absorbed pump power increased. Therefore, at the absorbed pump power ranging in our experiment up to 2 W, we concluded that the slight difference in thermal load among the three pump wavelengths does not affect the laser performance in our experiment.
Table 1 summarizes the Findlay–Clay analysis and the Caird analysis together with the absorption coefficients measured separately for the three pumping wavelengths. In spite of the aforementioned flawed analysis, fairly linear relations were achieved in these Findlay–Clay plots and Caird plots for three pumping wavelengths. Therefore, we extracted the effective residual losses from the plots. The estimated residual losses with two analysis schemes did not necessarily agree well with each other. The absorption coefficients estimated by the Caird plots with slope efficiencies were much smaller than those estimated by the Findlay–Clay analysis for Crystal A, while the absorption estimated by the Caird plots was larger than that estimated by the Findlay–Clay analysis for Crystal B at 520-nm pumping. These disagreements would be caused by difficulties in estimating threshold pump powers and accuracies in determining slope efficiencies. Since we obtained only four plots corresponding to different output couplings for the Findlay–Clay and the Caird analyses, the accuracy of the estimated pump-induced loss coefficients may not be so high. However, it is obvious that the intracavity loss increases as the pump wavelength becomes shorter; this is the major cause for the degradation in the laser slope efficiency. The rate-of-loss increase as the pump wavelength becomes shorter is different for each crystal. We calculated the effective FOMs defined by the ratio of absorption coefficients at the pump wavelength and the lasing wavelength during lasing; they became far removed from those specified by the manufacturers based on the spectroscopic measurements at 514 nm and 800 nm for each crystal. The effective FOM directly relates to a gain-to-loss ratio and determines the laser extraction efficiency. It is very clear that the effective FOMs drastically decrease from 520- to 451-nm pumping for the three crystals.
With the same analysis method, Roth et al. estimated that cavity loss increased at when a 5-mm-long Ti:sapphire crystal was pumped by a 458-nm LD . Therefore, our experimental results qualitatively agree with their observation.
C. Speculation of Mechanism of Pump-induced Loss
From Roth’s experiment  and our experimental evidence, a pump-induced 790-nm loss clearly exists at a pumping wavelength of around 450 nm. A similar phenomenon has already been reported at 464 nm in a flash-pumped Ti:sapphire laser by Hoffstädt , who explained this phenomenon by the charge transfer between and in the Ti:sapphire crystal. The excited level (the upper laser level) of locates at 2.01 eV above the ground level. If excited-state absorption takes place at 464-nm (2.67-eV) pump light, the charge transfer from excited-state to could be induced because its threshold from ground-states to is photon energy . Then, IR absorption induced by pairs appears around 800 nm [9,11]. Wong et al. reported that the threshold energy for the charge transfer from ground-states to is , consistent with Hoffstädt’s estimation. Wong et al. also indicated that the threshold energy for a charge transfer from ground-states to is 4.17 eV .
By assuming that this charge transfer is also induced by 451-nm (2.75-eV) LD light, we constructed the following mechanism to explain the laser power response at dual-wavelength pumping in Figs. 2–6. Figure 9 shows a schematic energy diagram of and . 451-nm light can induce a charge transfer from excited-state to . There could also be reverse charge transfers from to induced by 451-nm light. Since these reversible processes can quickly meet a steady-state balance, we could not observe any transient change in the laser output power when only the 451-nm LD was used for pumping (Fig. 5). Presumably, Durfee et al.  could not observe any laser power degradation at the blue-only pumping due to the fast-reversible reactions. On the other hand, since the 520-nm (2.39-eV) laser can recover its power (Fig. 3), the 520-nm light must also induce a reverse charge transfer from to . However, the photon energy of the green light is insufficient to induce a charge transfer from excited-state to . Therefore, a relatively slow reverse charge transfer rate of to by green light retards the degradation of the laser power at dual-wavelength pumping with green and 451-nm pumping. The photon energy of the 478-nm (2.59-eV) light is probably close to the threshold but remains sufficient to induce charge transfers to . We repeated the same experiment as in Fig. 3 by replacing the 521-nm LDs by the 478-nm LD and observed similar slow power reduction after termination of the 451-nm LD pumping. Therefore, 478- and 521-nm LDs principally have the same role in reverse charge-transfer induction. We did not observe a slow power decrease during the measurement with 478-nm pumping in Fig. 6 because presumably both the forward and reverse charge transfer reactions by 478 nm take place at similar slow rates as that of the reverse reaction by 520-nm light. Thus, reversible processes quickly meet a steady-state balance.
Based on this hypothesis on a pump-induced loss mechanism with charge transfer reactions, the excited-state density of ions at 451-nm pumping are determined by the steady-state condition achieved by the forward and reverse reactions. Therefore, a ground-state ion concentration at the crystal fabrication will not directly affect the steady-state excited-state pair density during 451-nm pumping. From Fig. 8 and Table 1, in fact, the output performances at 451-nm pumping were almost the same among three crystals, although the laser performance at 520-nm pumping was significantly different. Therefore, we emphasize that the influence of the pump-induced loss at 451-nm pumping is significant even for a crystal that exhibits higher effective FOM and excellent laser performance at 520-nm pumping.
3. Ti:sapphire LASER PUMPED BY WAVELENGTH MULTIPLEXED 520- AND 478-nm InGaN LASER DIODES
A. Experimental Setup
Since the present green LD’s power remains limited to 1 W, some beam-combining technologies are necessary for scaling pump powers and increasing the Ti:sapphire laser output power. Therefore, we demonstrated wavelength multiplexed pumping with a 520-nm LD pair and a 478-nm LD pair using the same experimental setup shown in Fig. 7. We used Crystals A and C. The dichroic pump mirrors have a radius of curvature of 75 mm. As shown in Fig. 7, the cavity was in an X-folded layout consisting of an output coupler and a high reflective mirror. The cavity length was 50 cm. The fold angles at the dichroic pump mirrors were set at 8° based on astigmatism compensation.
At the experiment using the Crystal C (), the highest total absorbed pump power in the Ti:sapphire crystal was 2.8 W. The calculated pump beam spot radius in the Ti:sapphire crystal were () for the 520-nm LDs and for the 478-nm LDs. The beam spot radius of the cavity mode was . The mode-matching efficiencies between the pump beams and the cavity mode were estimated to be for the 520-nm pumping and for the 478-nm LD pumping for this cavity design.
With the crystal A (), the highest total absorbed pump power in the Ti:sapphire crystal was 2.5 W. The calculated pump beam spot radius in the Ti:sapphire crystal were () for both the 520- and 478-nm LDs. The beam spot radius of the cavity mode was . The mode-matching efficiencies were approximately the same as those of Crystal C.
B. Results and Discussion
Figures 10 and 11 show the output power of Ti:sapphire lasers pumped with green (520 nm) and blue (478 nm) LDs in CW operation with Crystals C and A, respectively. For acquiring the dependence of the output laser power at dual-wavelength pumping, we adjusted the pumping optics so that the highest output power was obtained at the highest dual-wavelength pumping, lowered the applied current of the 478-nm LDs one by one, and then lowered the 520-nm LDs. During this process, no additional cavity alignment was adjusted.
Using the 4-mm-long Crystal C (Fig. 10), we observed increases of the threshold and the slope efficiency when changing to a higher transmission of output couplers and consequently obtained a maximum output power of 370 mW with an output coupling of 6.3%. Comparing the characteristics of CW laser operation for each pump wavelength, the slope efficiency at 520-nm pumping is always better than 478-nm pumping, partly because of a quantum defect (520/790 versus 478/790) since the mode matching efficiencies are comparable. While lowering the Ti:sapphire crystal, the cavity condition gradually departed from the optimum as the 478-nm pumping power decreased. With the pumping power of 478 nm, the slope efficiency gradually decreased. Presumably, since we optimized only the cavity and pumping optics at the highest pumping power where the thermal load formed a lens in the Ti:sapphire crystal, the cavity condition gradually departed from the optimum as the 478-nm pumping power decreased.
When we simply replaced Crystal C with Crystal A, the improvement in the output laser power was significant; this was already shown in Fig. 8, especially at 520-nm pumping. We obtained the highest output power of 593 mW at an absorbed pump power of 2.5 W. We obtained a laser conversion efficiency from the absorbed pump power to 790-nm laser power of 24%, significantly outperforming that obtained with 451-nm LD pumping. To explain such a difference in the laser performance where the slope efficiency of Crystal A was more than twice as high as that of Crystal C, the effective FOMs estimated by Caird analysis, 328 and 190 for Crystals A and C at 520-nm pumping, seem to be an adequate explanation.
We experimentally revealed the existence of a pump-induced loss for 451-nm pump wavelengths and explained the wavelength-dependent transient laser output power changes by charge transfer reactions induced by pumping lasers. Also, we found that the pump-induced loss was reduced for a pump wavelength of 478 nm. Note that the influence of the pump-induced loss at 451-nm pumping is significant even for a crystal that exhibits higher effective FOM and excellent laser performance at 520-nm pumping.
We proposed an efficient wavelength multiplexed pumping scheme using two green (520-nm) LDs and two blue (478-nm) LDs, and obtained a highest Ti:sapphire laser power of 593 mW at a total absorbed pump power of 2.5 W.
Ministry of Education, Culture, Sports, Science, and Technology (MEXT) (26286065).
The authors thank the NICHIA Co. for providing high-power InGaN blue and green LDs.
1. P. Roth, A. Maclean, D. Burns, and A. Kemp, “Directly diode-laser-pumped Ti:sapphire laser,” Opt. Lett. 34, 3334–3336 (2009). [CrossRef]
2. P. Roth, A. Maclean, D. Burns, and A. Kemp, “Direct diode-laser pumping of a mode-locked Ti:sapphire laser,” Opt. Lett. 36, 304–306 (2011). [CrossRef]
3. M. D. Young, S. Backus, C. Durfee, and J. Squier, “Multiphoton imaging with a direct-diode pumped femtosecond Ti:sapphire laser,” J. Microsc. 249, 83–86 (2013). [CrossRef]
4. S. Sawai, A. Hosaka, H. Kawauchi, K. Hirosawa, and F. Kannari, “Demonstration of a Ti:sapphire mode-locked laser pumped directly with a green diode laser,” Appl. Phys. Express 7, 022702 (2014). [CrossRef]
5. K. Gürel, V. J. Wittwer, M. Hoffmann, C. J. Saraceno, S. Hakobyan, B. Resan, A. Rohrbacher, K. Weingarten, S. Schilt, and T. Südmeyer, “Green-diode-pumped femtosecond Ti:sapphire laser with up to 450 mW average power,” Opt. Express 23, 30043–30048 (2015). [CrossRef]
6. C. G. Durfee, T. Storz, J. Garlick, S. Hill, J. A. Squier, M. Kirchner, G. Taft, K. Shea, H. Kapteyn, M. Murnane, and S. Backus, “Direct diode-pumped Kerr-lens mode-locked Ti:sapphire laser,” Opt. Express 20, 13677–13683 (2012). [CrossRef]
7. D. Findlay and R. A. Clay, “The measurement of internal losses in 4-level lasers,” Phys. Lett. 20, 277–278 (1966). [CrossRef]
8. J. A. Caird, S. A. Payne, P. R. Staver, A. J. Ramponi, L. L. Chase, and W. F. Krupke, “Quantum electronic properties of the Na3Ga2Li3F12:Cr3+ laser,” IEEE J. Quantum Electron. 24, 1077–1099 (1988). [CrossRef]
9. A. Hoffstadt, “Design and performance of a high-average-power flashlamp-pumped Ti:sapphire laser and amplifier,” IEEE J. Quantum Electron. 33, 1850–1863 (1997). [CrossRef]
10. N. A. Moskvin, V. A. Sandulenko, and E. A. Sidorova, “Color centers and luminescence in corundum crystals containing titanium,” J. Appl. Spectrosc. 32, 592–596 (1980). [CrossRef]
11. A. Sanchez, A. J. Strauss, R. L. Aggarwal, and R. E. Fahey, “Crystal growth, spectroscopy, and laser characteristics of Ti:Al2O3,” IEEE J. Quantum Electron. 24, 995–1002 (1988). [CrossRef]
12. W. C. Wong, D. S. McClure, S. A. Basun, and M. R. Kokta, “Charge-exchange processes in titanium-doped sapphire crystals. I. Charge-exchange energies and titanium-bound excitations,” Phys. Rev. B51, 5683–5692 (1995).