Femtosecond CPA hybrid laser system with pulse-on-demand operation

: In this manuscript we present a true pulse-on-demand concept of a hybrid CPA laser system, consisting of a chirped-pulse ﬁber ampliﬁer and an additional solid-state ampliﬁer, capable of generating femtosecond pulses on demand without an external optical modulator/shutter. Pulse-on-demand operation is achieved by introducing idler pulses with a few nanoseconds duration and selectively switching between the femtosecond and idler pulses. The idler pulses are used to maintain a constant population inversion in the ﬁber ampliﬁer as well as in the solid-state ampliﬁer. Second harmonic generation (SHG) unit then eﬀectively ﬁlters out the idler pulses due to their low peak power, leaving only a stable femtosecond pulse train. This concept is demonstrated on a CPA hybrid system that can generate pulses with up to 200 µJ at 515 nm with a pulse duration under 450 fs. As there is no optical modulator at the laser output, the presented concept also enables further power scaling.


Introduction
In industrial applications the speed of manufacturing is of great importance. To achieve high speed laser processing of materials, polygon scanners or resonant scanners must be used [1,2] in combination with high power lasers with high pulse repetition rates to achieve a sufficient processing resolution. Due to the fixed scanning speeds of such scanners, lasers with a so-called pulse on demand operation must be used to enable accurate positioning of the laser pulses on the sample being processed.
Different methods for the pulse on demand operation have been reported. One possible method that offers high control over pulse generation in the nanosecond range is the gain-switching approach [3][4][5]. However, numerous new applications require ultrashort pulses therefore this method is unsuitable.
With the development of coherent beam combination techniques, a different approach based on precise phase control between the combined beams can also be used, achieving pulse on demand operation by switching between the constructive and destructive interference of the combined beams [6].
Other methods to achieve pulse on demand operation usually include an external pulse-picking device, either an electro-optical modulator (EOM) or acousto-optical modulator (AOM) [7] to cut out the desired pulse pattern from a fixed repetition rate laser [8,9] In the case of AOM, tight focusing is required to reach high modulation speeds. The modulation speed is consequently inversely proportional to the laser power due to the damage threshold limit in the AOM. Similarly, in the case of EOM, the laser power is limited by the onset of thermal lensing caused by significant absorption in the EOM crystal. External pulse-picking is therefore unsuitable for high-power high-repetition rate lasers.
Another method to produce pulses at variable repetition rates is the so-called burst operation mode [10][11][12][13][14][15][16]. In this case the pulses are grouped into bursts and complex burst pre-shaping and pump power modulation is required to counteract the gain saturation and prevent excessive gain buildup respectively, ultimately producing homogeneous bursts of pulses at the output of the amplifier. Although arbitrary burst repetition rates and pulse counts within the bursts are easily selectable, they are usually kept fixed at least on the time scale of the laser scanning. Producing truly arbitrary pulse sequences, i.e. with continuously variable pulse to pulse periods would require an even more complex burst-preshaping envelopes and pump modulation schemes.
The solution for the above problem is to control the gain in the laser amplifiers in order to eliminate transient pulse dynamics that appear as a direct consequence of non-uniform pulse distributions.
To achieve gain control in pulse on demand laser systems, several approaches can be used. For example one can modulate the pump power [17] or use an additional idler source that keeps the population inversion at a constant level [18,19]. In the latter case the idler and the primary seed pulses have to be separated after the amplification. To do so, one can use different wavelengths or different polarizations for the idler and primary seed pulses and then separate them with an optical filter or a polarization beam splitter respectively [19]. Another approach is to introduce a CW idler or idler pulses that have a relatively long duration (compared to the primary seed pulses) and which can be separated from the primary seed pulses via the second harmonic generation (SHG) at the laser output [18,20].
Additional obstacle that has to be overcome with high energy ultrashort pulses are nonlinear effects in optical fibers [21][22][23][24][25][26]. To achieve high energies in fiber amplifiers, complex and expensive solutions must be applied [27,28]. Alternatively a simpler and more compact laser that enables relatively high pulse energies is a so called hybrid laser system, which implements an additional solid-state amplifier after the fiber pre-amplifier [14,29,30]. Due to the larger mode field area and shorter interaction length, the thresholds for nonlinear effects are several orders of magnitude higher in the solid-state amplifier and therefore higher pulse energies can be achieved.
Although hybrid laser systems allow higher pulse energies with less nonlinear effects, a pulse on demand operation with ultrashort pulses in such systems is nontrivial. Even with an additional idler source for gain control, the idler and ultrashort laser pulses usually exhibit different gain in the fiber and solid-state amplifiers due to the different gain spectra of the fiber and solid-state amplifiers. In particular, Yb-doped fiber amplifiers have a broad gain spectrum, consequently both the ultrashort laser pulses with a bandwidth of several nanometers as well as the narrowband idler pulses exhibit the same gain. However, in an Yb:YAG solid-state amplifier, the idler pulses exhibit higher gain when compared to the ultrashort pulses due to the narrow gain bandwidth of the Yb:YAG crystal. Compensating the gain mismatch between the fiber and solid-state amplifiers would require complex amplitude modulations of both the ultrashort and idler pulses.
In this manuscript we present a hybrid laser system, which produces high-energy ultrashort pulses on demand by introducing idler pulses with a duration of several nanoseconds for gain control in the amplifiers. We propose and experimentally evaluate spectral tuning of the wavelength of the idler pulses in order to compensate for the gain mismatch between the fiber and solid-state amplifiers and achieve a true pulse on demand operation.
The manuscript is organized as follows: first the theory behind pulse amplification that considers different gain for ultrashort and idler pulses in the fiber and solid-state amplifier is introduced with detailed insight to the problem of pulse on demand operation in a hybrid laser system. The solution to the mentioned problem is also presented in this chapter. Second, the experimental setup used to obtain the presented results is described followed by experimental results and their analysis. Finally, a summary of the work is given in the conclusion.

Theory
To achieve a true pulse on demand operation based on the concept of gain control, the primary (femtosecond) and the idler (nanosecond) seed pulses need to have the same gain in all amplifier stages to eliminate transient pulse dynamics when switching between idler and primary seed pulses. In a fiber laser system this is achieved by carefully tuning the idler pulse energies to exactly match those of the femtosecond pulses. In a hybrid laser however, there is an additional gain-spectra mismatch between the fiber and the solid-state amplifier that must be considered. Yb-doped fibers have a broad gain spectrum whereas the gain spectrum of an Yb:YAG solid-state amplifier is relatively narrow. This inevitably results in a gain mismatch between the typical femtosecond pulses with a broad spectrum (FWHM of 6.1 nm in our case) and the typical idler pulses generated by a laser diode that usually has a narrow spectrum. The measured primary seed, idler and Yb:YAG emission spectra are shown in Fig. 1. Because the primary seed spectrum (FWHM 6.1 nm) is broader than the solid-state gain spectrum, the primary seed pulses experience a lower gain than the idler pulses, which results in transient effects when switching between the primary seed and the idler pulses, as the population inversion of the solid-state amplifier slowly adapts to the different seeding conditions. The transient effects are schematically shown in Fig. 2. Because the gain of the fiber amplifier is nearly constant over the whole spectrum of the primary seed pulses, this effect is not observed at the output of the fiber amplifier as seen in Fig. 2. The gain in the fiber amplifier is the same for both the primary seed and idler pulses and therefore the fiber amplifier operates at constant seeding conditions and no transient pulse dynamics occur.  To compensate for the different gain spectra, a method of wavelength tuning (by temperature) of the idler laser diode to achieve the same average gain for idler and primary seed pulses in the solid-state amplifier is presented and demonstrated.

Pulse amplification
Pulse propagation in laser amplifiers can be modeled using the Frantz-Nodvik equations [31] which model the photon density and population inversion as a function of time and position in the laser amplifier. In our case we are interested only in the pulse energy at the output of the amplifier. In this case the Frantz-Nodvik equations can be simplified in a way that temporal and longitudinal dependencies are averaged out. The simplified equations are as follows: where E out is the pulse energy at the output of the amplifier, E s is the saturation energy of the amplifier, E in is the pulse energy at the input of the amplifier, ∆ out is the population inversion of the amplifier after the pulse leaves the amplifier, σ is the emission cross section and L is the length of the amplifier. Small signal gain g in at the time when laser pulse enters the amplifier is introduced as g in = exp(σL∆ in ), where ∆ in is the population inversion at the time that laser pulse enters the amplifier.
As the timescale of the pulse on demand operation is large in comparison to the pumping rate, the pumping of the gain media must be considered. The change in the population inversion due to pumping between the pulses can be modeled as where ∆ is the population inversion as a function of time after the pulse leaves the amplifier, A = 2σ p cn p − 1/τ and B = σ p cn p N, where σ p and n p are the absorption cross section for the pump photons and the density of the pump photons respectively. τ is the upper state lifetime of the active ions and N is the density of the active ions in the gain medium. Pumping is neglected while the pulse is in the amplifier.
To model pulse amplification in a hybrid laser, Eqs. (1) and (2) were used. The model was divided into two parts. First, we modeled pulse amplification in the fiber amplifier followed by the amplification in the solid-state amplifier.
The pulse train consisted of the idler and primary seed pulses, with the same repetition rate and the same input energy. By experimentally measuring the gain difference for the primary seed pulses and the idler pulses in the solid-state amplifier, we introduced a gain mismatch factor of 0.6 to the primary seed pulses in our model. The results of the simulation and comparison with experimental results are shown in Fig. 3.
It can be seen from Fig. 3 that the idler and primary seed pulses both have the same gain in the fiber amplifier but exhibit different gain in the solid-state amplifier. Because the idler pulses experience a higher gain in the solid-state amplifier than the primary seed pulses, the conditions in the solid-state amplifier change with time. This causes transient pulse dynamics i.e. changes in pulse to pulse intensity in the pulse train after the solid-state amplifier. Due to the higher gain for the idler pulses in the solid-state amplifier, the amplitudes of the first primary seed pulses are lower compared to the steady state values. This is a consequence of lower population inversion (due to the higher gain) in the amplifier after the last idler pulse compared to the population inversion when only primary seed pulses are present in the amplifier. The amplitudes of the primary seed pulses then gradually increase with consecutive pulses as the amplifier slowly adapts to the new seeding conditions.

Temperature tuning of the idler wavelength
To achieve the same net gain in the solid-state amplifier for the idler and primary seed pulses one possibility is to shift the wavelength of the idler diode away from the center of the solid-state Results are shown at the output of the fiber stage denoted with "Fiber" and at the output of the solid-state amplifier denoted by "Solid state". The idler pulses are colored with red and primary seed pulses with blue. The lower amplitude of the idler pulses is a consequence of longer pulse durations of idler pulses which results in lower peak power at the same pulse energy. It should also be noted that primary seed pulse peak powers are not to scale due to the limited rise-time of the photodiode used (a few ns). gain curve. This can be achieved with temperature tuning of the idler laser diode. By shifting the wavelength away from the center, the gain for the idler pulses in solid-state amplifier reduces but remains approximately the same in the fiber amplifier due to the broad gain spectrum of the fiber amplifier. Using the Frantz-Nodvik Eq. (1), we calculated the emission cross-section of the idler pulses, that result in the same gain as for the primary seed pulses. To achieve the same gain for the primary seed and idler pulses, the effective emission cross section for the idler had to be reduced to 77% of the maximum emission cross section at 1030 nm. Since the emission cross section is proportional to the emission spectrum, we can see from Fig. 4 that in our case the wavelength of the idler diode has to be shifted either towards 1031.5 nm or towards 1028.7 nm to achieve the desired cross-section. As mentioned before the shift in the wavelength was achieved by temperature tuning of the idler diode. The spectra of the idler diode at different temperatures are shown in Fig. 4.

Experimental setup
The experimental setup consists of a fiber pre-amplifier and an additional solid-state amplifier and is shown in Fig. 5. The seed laser is a commercial mode locked fiber oscillator that generates 1.5 ps pulses at 30 MHz repetition rate and 1 mW average power at 1030 nm. Pulse picker is used to pick primary seed pulses from the 30 MHz pulse train. The pulses are then amplified in a system consisting of three fiber amplifiers and finally in an additional Yb:YAG solid-state amplifier. The fiber amplifiers are pumped with 976 nm diode lasers. To avoid the nonlinear effects, the primary seed pulses are temporally stretched to 500 ps using a tunable chirped fiber Bragg grating. At the output of the solid-state amplifier the pulses are compressed using a transmission grating compressor to 450 fs with an ∼80% compression efficiency. The solid-state Fig. 5. The experimental setup, consisting of a front end (a), consisting of a mode locked fiber oscillator generating picosecond pulses at 30 MHz repetition rate, pulse stretcher (chirped fiber Bragg grating), fiber amplifier and pulse picker (P. P.). Pulses from the primary seed module are then amplified in two additional fiber amplifiers (c) and (d) and in one additional Yb:YAG amplifier (e) before being recompressed in a transmissive grating compressor (f). To achieve pulse on demand operation additional idler pulses from a gain-switched laser diode, with 32 ns pulse duration (b) are introduced after the primary seed module, replacing individual primary seed pulses. Control electronics are used to control the pulse picker and idler diode and to construct arbitrary pulse on demand sequences. Finally, a second harmonic generation (SHG) unit is used to filter-out the long idler pulses from the laser output (g). amplifier used was an Yb:YAG single-crystal amplifier module pumped with two 969 nm diode lasers with a total power of 200 W and optical-to-optical efficiency up to 25%. With the described setup we achieved 300 µJ pulses at 1030 nm with 450 fs pulse duration at 100 kHz repetition rate with M 2 value of 1.3 in the x direction and 1.5 in the y direction. The measured autocorrelation trace of the pulse is shown in Fig. 6. To achieve the pulse on demand operation, additional idler pulses are introduced after the pulse picker. The idler pulses are generated using a 1030 nm DFB laser diode. The pulse duration of the idler pulses was set to 32 ns. The energy of the idler pulses was matched to that of the primary seed pulses, to maintain a constant seeding power of the amplifiers when switching between the idler and primary seed pulses. Control electronics were used to control the pulse picker and to switch between the primary seed and idler pulses in order to achieve the pulse on demand operation. A schematic of pulse on demand operation is shown in Fig. 7. The laser system is designed to operate at an internal repetition rate up to 2 MHz. We use the term internal repetition rate to describe the repetition rate of the combined primary and idler pulses within the laser amplifiers chain, i.e. before the idler-seed separation. The external repetition rate of the primary seed pulses alone can therefore vary from single pulse up to the internal repetition rate. The temporal resolution however is 30 MHz and is limited by the mode locked seed's repetition rate. The internal pulse repetition rate can be changed in order to meet for example, the scanning speed and pulse energy requirements. In this manuscript two internal repetition rates are analyzed: 2 MHz and 100 kHz. The last one also corresponds to the internal repetition rate at which the maximum pulse energy was achieved.
To separate the idler and the primary seed pulses at the output of the compressor, second harmonic generation is used. For this we used a 2 mm long LBO crystal at 55°C. As the primary seed pulses are compressed to 450 fs (having negligible effect on the idler pulses) this results in a high peak-power contrast between the seed and idler pulses. Because of the low peak power and consequently the low conversion efficiency of the idler pulses, this results in a very high contrast ratio of 1:3000 in pulse energy (more than 1: 10 8 in peak power) between the primary seed and idler pulses at 515 nm.
The laser system can produce pulses on demand with a compressed pulse energy up to 200 µJ at 515 nm (100 kHz internal repetition rate) with a SHG efficiency of 67%. The SHG efficiency as high as 80% was achieved at the internal pulse repetition rate of 2 MHz where the achieved Fig. 7. A schematic of the pulse on demand method used. The function of the pulse picker (P. P.) is twofold. Firstly, it reduces the repetition rate of the primary seed pulses from 30 MHz to an internal repetition rate (2 MHz and 100 kHz presented in this manuscript) in order to meet for example, the scanning speed and pulse energy requirements. Secondly the pulse picker generates an arbitrary primary seed pulse sequence, i.e. pulse on demand sequence. Due to the high repetition rate of the primary seed pulse sequence before the P. P., a temporal resolution of 30 MHz can be achieved in the pulse on demand trace after the pulse picker. To replace the missing primary seed pulses after the pulse picker, idler pulses (red) are introduced in order to maintain constant seeding conditions for the amplifier stages in the pulse on demand operation. pulse energy was 20 µJ at 515 nm. At 100 kHz internal repetition rate the conversion efficiency decreased as the laser spot size in the SHG crystal was increased in order to avoid optical damage.
The pulse duration at 515 nm was not measured but is expected to be shorter than 450 fs as measured at 1030 nm.

Results and discussion
As presented in section 2 the wavelength of the idler diode can be tuned for more than 1 nm. The calculated wavelength of the idler that results in the same gain for the idler and primary seed pulses in the solid-state amplifier is 1031.5 nm. By measuring the temperature dependence of the idler spectrum, we determined that the idler diode must be heated to 45°C. Pulse on demand results with a heated idler diode are shown in Fig. 8. The results shown in Fig. 8(a) to Fig. 8(d) are measured at 100 kHz internal repetition rate and the results shown in Fig. 8(e) to Fig. 8(h) are measured at 2 MHz internal repetition rate.
In Fig. 8(b) and Fig. 8(f) transient pulse dynamics can be observed after the solid-state amplifier when switching from the idler to primary seed pulses and vice versa. The transient effects after the solid-state amplifier are significantly reduced at higher idler diode temperature which can be seen in Fig. 8(d) and Fig. 8(h) as the gain of the idler in the solid-state amplifier is reduced. It can be further observed in Fig. 8 that wavelength tuning of the idler diode has no effect on the fiber amplifier, as seen in Fig. 8(c) and Fig. 8(g), where no transient effects are observed at the output of the fiber amplifier even with the heated idler diode.
Although transient effects are reduced with heated idler diode, some residual transient effects can be observed in Fig. 8(h) at 2 MHz internal repetition rate whereas the transient effects are almost completely eliminated at 100 kHz internal repetition rate shown in Fig. 8(d). We attribute this effect to additional heating of the idler diode due to the driving current. The magnitude of residual transient effects varies with idler sequence duration and idler diode duty cycle. The problem is more clearly shown in Fig. 9 where special pulse on demand sequence is shown in which the above problem is emphasized.
The pulse on demand sequence in Fig. 9, consists of variable primary seed pulse sequences followed by variable idler pulse sequences (i.e. variable duty cycle). When comparing Fig. 9(b) and Fig. 9(d) we can notice that no transient pulse dynamics can be observed at the output of the solid-state amplifier at 100 kHz internal repetition rate, whereas at 2 MHz internal repetition rate, the amplitude of the primary seed pulses varies significantly along the primary seed pulse sequence. It can be further observed that the magnitude of the transient pulse dynamics depends on the interval between the primary seed sequences. As already mentioned above, this effect can be attributed to the additional heating of the idler diode due to the driving current. Because the idler diode does not operate at a constant heat load, the effect of the heating depends on the duration of the idler sequence and the duration of the primary seed pulse sequence. Because the duty cycle of the idler diode is much higher at 2 MHz (i.e. ∼ 10%) the effect is only seen at higher repetition rates. At 100 kHz the duty cycle is ∼ 0.5% therefore the effect of the heating is negligible.
In order to eliminate the transient dynamics at higher repetition rates the idler diode must be kept under a constant heat load. This can be achieved by applying a constant current that is just below the lasing threshold during the primary seed pulse sequence when the idler pulses are switched off. The results obtained with this method are shown in Fig. 10.   Fig. 10. Pulse on demand sequence with 2 MHz internal repetition rate and with a constant load on the idler diode. During the primary seed pulse sequences, a constant current (below the laser threshold) was applied to the idler diode, consequently achieving a constant heat load on the idler diode. The results at the output of the fiber amplifier are denoted with "Fiber" and at the output of the solid-state amplifier with "Solid state".
To generate the idler pulses, 150 mA of current was applied to the idler diode. To keep the idler diode under a constant heat load, the idler diode current during the primary seed pulses needed to be 10 mA in the case of 2 MHz internal repetition rate and only 0.5 mA in the case of 100 kHz internal repetition rate. In both cases, the current is well below the idler diode threshold current (30 mA). The idler diode current in the idler OFF state can be calculated as where I off is the current in the idler OFF state, I on is the current while idler pulses are ON, τ on is the duration of the idler pulses which is in our case 32 ns and τ off is the length of idler OFF state and it depends on the internal repetition frequency. In the case of 2 MHz internal repetition rate, the τ off = 500 ns and in the case of 100 kHz internal repetition rate the τ off = 10 µs. The idler OFF current can be neglected in the case of 100 kHz internal repetition rate as shown in Fig. 9(b).
As we can see in Fig. 10(b) this method efficiently eliminates transient pulse dynamics after the solid-state amplifier. Using this method, we can achieve equal pulse energies in all the primary seed pulses, which is necessary for a true pulse on demand operation.

Primary seed and idler separation
To separate the idler pulses from the primary seed pulses, a second harmonic generation (SHG) was used. The SHG crystal used was a 2 mm long LBO crystal at 55°C temperature. The focal spot in the crystal was approximately 100 µm in diameter and was optimized with a variable beam expander to achieve optimal conversion to 515 nm.
The efficiency of the SHG for the long idler pulses was significantly lower than the efficiency for the primary (femtosecond) seed pulses. The conversion efficiency for the primary seed pulses was 80% at low pulse energies and 67% at high pulse energies as the focal spot in the SHG crystal Fig. 11. Comparison of the pulse on demand sequences after the fiber amplifier (a), solid-state amplifier (b) and after the second harmonic generation (c). was increased in order to avoid optical damage. The contrast between the idler and primary seed pulses after the SHG was 1:3000 in pulse energy (more than 1: 10 8 in peak power). The contrast between the primary seed and idler pulses was determined by measuring the SHG power of the signal consisting only of the primary seed pulses and separately the SHG power of the signal consisting only of the idler pulses. As the average power prior to the SHG was the same for both signals, the contrast was determined by the ratio of average power between the two signals after the SHG.
By tuning the idler wavelength, we achieved excellent results in pulse on demand operation with no transient pulse dynamics. The pulse on demand results after the second harmonic generation are shown in Fig. 11.
The results in Fig. 11 show a true pulse on demand operation with no observable transient pulse dynamics. The interval between the primary seed pulse sequences does not affect the primary seed pulse energies after the SHG. With the described setup we can therefore construct an arbitrary primary seed pulse sequence with idler pulses replacing individual primary seed pulses for gain control, and consequently achieve equal energies of all the primary seed pulses after both the fiber and solid-state amplifiers and also after the SHG.

Conclusion
To achieve a true pulse on demand operation in a high-power laser all output pulses need to have equal energies. To achieve this and therefore eliminate any transient pulse dynamics in a pulse on demand system, some sort of gain control mechanism needs to be implemented. Special care must be taken with gain control in a hybrid laser system i.e. when combining fiber amplifier with a solid-state amplifier as the latter usually has a narrower gain spectrum compared to the first.
In this work, we investigated gain control by introducing the idler pulses to replace the primary seed pulses when needed in order to keep the gain in the amplifiers chain constant. To compensate for the different gain spectra of the fiber and the solid-state amplifiers a method of temperature tuning of the idler wavelength was presented. In order to adjust the net idler gain in the solid-state amplifier while maintaining the same gain in the fiber amplifier idler wavelength was shifted away from the maximum of the solid-state gain curve.
Using the above-mentioned approach, complex pulse on demand sequences with no transient pulse dynamics were achieved.
To separate the idler pulses from the primary seed pulses at the output of the amplifiers SHG was used. Using this technique, a contrast ratio between the idler and primary seed pulses of 1:3000 was achieved at 67%−80% conversion efficiency for the primary seed pulses.
We achieved transient free pulse on demand operation at 515 nm with pulse energies up to 200 µJ.