1. Introduction

Ultrafast fiber lasers (UFLs), generating short optical pulses in picosecond and femtosecond ranges, have become a powerful tool in industrial and medical fields as well as for scientific researches [1–8]. Particularly, high-repetition-rate (HRR) UFLs have been garnering increasing research interests due to their new potentials in various fields, such as optical frequency combs, nonlinear bioimaging and optical communications [9–11]. Currently, the widely adopted strategy for the generation of HRR ultrafast pulses from fiberized cavities is passive mode-locking—a technique based on nonlinear optical phenomena (e.g., saturable absorption, nonlinear polarization rotation and nonlinear amplification loop mirror), which possesses a lot of prominent merits such as self-starting operation, low cost, high stability and free of maintenance [12–20]. In general, the fundamental repetition rate of a passively mode-locked fiber laser is reversely related to the total cavity length [9]. In order to generate HRR pulses, thus, the active fiber is required to provide a high gain for shortening the cavity length. The past decades have witnessed great progress in development and fabrication of various active fibers, especially, the Yb³⁺-doped, Er³⁺-doped and Tm³⁺-doped fibers. By adopting multi-component glass as the host, the fiber gain coefficients have been boosted to 5.7 dB cm⁻¹ at 1.0 µm [21], 5.2 dB cm⁻¹ at 1535 nm [22] and 3.6 dB cm⁻¹ at 1950 nm [23], which enable the construction of ultra-short mode-locking cavity in centimeter long, giving rise to a fundamental repetition rate on the order of GHz.

Serving as laser sources in biophotonic systems, UFLs at 0.9 µm has attracted intensive attention due to two significant advantages. First, widely-used fluorescent proteins, especially the enhanced green fluorescent protein (EGFP), present stronger two-photon excitation at this band, therefore, the optical imaging systems based on 0.9-µm UFL could be applied in deep tissue imaging and in vivo dynamic observation [24,25]; Secondly, compared to traditional solid-state lasers, e.g., Ti:sapphire lasers, fiber lasers are superior in terms of beam quality, compactness and robustness and could enhance the reliability, portability and cost-efficiency of bio-imaging systems, which is beneficial for their practically clinical applications [25–27].

So far, two major approaches have been used to realize UFL at 0.9 µm, i.e., direct generation based on Nd³⁺-doped optical fibers [24,27–29]; and nonlinear frequency conversion [26,30]. Among them, the direct generation using Nd³⁺-doped fibers is a desirable way on account of its simplicity and high stability, because the three-level transition (⁴F_3/2→⁴I_9/2) of Nd³⁺ exactly falls in the vicinity of 900 nm. As a consequence, considerable efforts have been devoted to exploring Nd³⁺-doped UFLs at 0.9 µm. Especially, the UFLs at 0.9 µm with high fundamental repetition rate have attracted a lot of attention due to the advantage in decreasing the photobleaching of fluorescent specimens [31]. However, owing to the poor gain of Nd³⁺-doped fiber at 0.9 µm caused by the problem of ground state absorption (GSA) and the competition with four-level transition (⁴F_3/2→⁴I_11/2) at 1.06 µm, the pulse repetition rate, operating at repetition rates ranging from a few to tens of MHz, is largely limited because the use of meters-long Nd³⁺-doped fiber to provide sufficient gain for laser oscillation. Therefore, developing high gain Nd³⁺-doped glass fibers is the key that lying at the heart of generating HRR UFLs at 0.9 µm.

Recently, Fu et al. realized a continuous wave 880 nm fiber laser based on Nd³⁺-doped phosphate glass fiber with superior conversion efficiency, and pointed out Nd³⁺-doped phosphate glass could be an efficient material for lasing at this band [32]. In this work, taking multi-component phosphate glass as the host, we design and fabricate a new Nd³⁺-doped single-mode fiber that provide a high gain at 0.9 µm. By introducing CCl₄ to reduce OH^- content during glass melting process, the luminescence efficiency of Nd³⁺-doped phosphate glass is prominently enhanced, leading to the enhancement fiber gain at 0.9 µm. Based on the Rod-in-Tube method, the fiber with a NA of 0.12 and a core diameter of 5 µm is successfully drawn. Compared to Nd³⁺-doped silica fiber, the gain coefficient of this fiber is improved from 1 dB cm⁻¹ to 2.7 dB cm⁻¹ at 915 nm, such that the slope efficiency of the continuous-wave (CW) fiber laser constructed with this fiber can be increased from 1.9% to 11.2%. By using this high gain fiber, we build a 0.9-µm passively mode-locked fiber laser with a centimeter-level cavity length for the first time, and it operates at a fundamental repetition rate of up to 1.2 GHz, a value two orders of magnitude higher than that reported in previous works [24,27–29]. We anticipate that this phosphate glass fiber with high gain at 0.9 µm could be a promising gain medium for fabricating advanced fiber lasers that are urgently demanded for multiphoton biophotonics.

2. Experimental section

2.1 Fiber fabrication

The glass samples with molar composition of 57.5SiO₂-16Al₂O₃-21ZnO-5TiO₂-0.5Nd₂O₃, 74.5GeO₂-10BaO-15Ga₂O₃-0.5Nd₂O₃, 62.5CaO-37Al₂O₃-0.5Nd₂O₃, 74.5B₂O₃-20BaO-5Al₂O₃-0.5Nd₂O₃ and 66.85P₂O₅-27.3(K₂O + BaO)-2.2Al₂O₃-3.15La₂O₃-0.5Nd₂O₃ were fabricated by conventional melting and quenching method. The 30 g batches of raw material were homogeneously mixed in an agate mortar and melted in an alumina crucible for 30 mins in air. Subsequently, the melts were quickly poured onto stainless steel plate and pressed with another steel plate to form solid sample. Then, the obtained glass samples were cut and polished into same thickness for optical and physical measurements.

To fabricate core bulk phosphate glass, the 200 g batch of raw material with above same phosphate composition was weighed and homogeneously mixed, and transferred into a covered alumina crucible and melted at 1250 °C for 2 hrs. The active dehydration agent of CCl₄ carried by dry O₂ was bubbled into melts during melting process to decrease the OH^- content and improve the glass homogeneity. Then, the melts were cast into a preheated steel mold and annealed near the glass transition temperature for 2 hrs and subsequently cooled to the room temperature at a rate of 0.05 °C min⁻¹. Another core bulk phosphate glass was fabricated by using similar procedures with only dry O₂ bubbling. The cladding bulk phosphate glass with the molar composition of 69.5P₂O₅-25.3(K₂O + BaO)-2.2Al₂O₃-3La₂O₃, in the quantity of 800 g, was melted with the same process of core glass.

For fiber fabrication, the obtained core bulk glass with incorporation of CCl₄ was mechanically ground and optically polished to a cylindrical rod with a diameter of 10 mm. Then, the core rod was suspended in a fiber drawing tower and drawn to the thin rod with a diameter of 1.6 mm. The cladding glass was directly cut, ground, drilled and polished to a tube with an inner diameter of 1.6 mm and an external diameter of 40.0 mm. The 1.6 mm core thin rod was inserted into the cladding tube to form a preform. To remove the surface contaminants, both the rod and tube were ultrasonically cleaned and etched by acid. The preform was suspended in a fiber drawing tower and drawn into single-mode fiber at a temperature of 720 °C. The fiber diameter was precisely controlled to a standard value of 125 ± 5 µm by adjusting the drawing speed and the feed speed of the preform.

2.2 Characterizations

The refractive index measurement of glass was conducted by a prism coupling equipment (Metricon Model 2010). Thermal properties of glasses were tested by differential scanning calorimetry (DSC) (STA449C NETZSCH) analyzer at a heating rate of 10 K min⁻¹ under nitrogen atmosphere. Near-infrared (NIR) emission spectra were taken on a Zolix Omni λ3007 spectrometer (Zolix Instruments, Inc., Beijing, China) equipped with an InGaAs photodetector (PD) and a SR830 Stanford Research lock-in amplifier. An 808 nm GaAlAs semiconductor laser diode (LD) was used as the excitation source. The Fourier Transform infrared spectroscopy (FTIR) was recorded on a Bruker Vertex 33 spectrometer. The elemental distribution of glass fiber was determined using an electro-probe micro-analyzer (EPMA) system (EPMA-1600, Shimadzu, Kyoto, Japan). The fluorescence decay curves of Nd³⁺:⁴F_3/2 level under excitation of a pulsed 808 nm LD were collected by a Tektronix TDS 3012c Digital Phosphor Oscilloscope. The laser properties were characterized on an optical spectrum analyzer (OSA, Yokogawa AQ6370) and a power meter (Ophir, 3A-P). The pulsed train was examined by a photodetector (PD) with a bandwidth of 12.5 GHz and collected by a real-time oscilloscope with a bandwidth of 20 GHz (Keysight DSOV204A). The repetition rate was analyzed by a phase noise analyzer (Rohde & Schwarz FSWP26, 26.5 GHz bandwidth). The pulse width is measured by an autocorrelator (APE Pulsecheck USB 50).

3. Results and discussion

3.1 Enhancement of luminescence efficiency of Nd³⁺-doped phosphate glass

The Nd³⁺ exhibits three fluorescent bands, i.e., 0.9, 1.06 and 1.3 µm, which originate from same upper energy level (⁴F_3/2) to different lower level ⁴I_J (J = 9/2, 11/2, 13/2) transitions, causing the strong competition among them, especially between 0.9 and 1.06 µm [33,34]. Generally, 0.9 µm is inferior to 1.06 µm in the competition because of the smaller branching ratio at 0.9 µm. In addition, ⁴F_3/2→⁴I_9/2 transition is a three-level structure and limited by ground state absorption. In our previous work, we have boosted the fluorescence branching ratio of 0.9 µm and strengthen the Stark splitting of ⁴I_9/2 level in Nd³⁺-doped multi-component phosphate glass by crystal field strength and structural engineering [33]. Compared to other glass systems, adopting phosphate glass as the host can largely increase the doping concentration of Rare-Earth (RE) ions, leading to the improvement of the gain coefficient. Despite the great opportunity, there is also an issue of high hydroxyl (OH^-) content due to strong hydroscopicity of phosphate, which degrades the performance of RE³⁺-doped phosphate glass fibers (and fiber lasers) from two aspects [35]. On the one hand, the OH^- could directly transfer the energy of excited RE ions, which quickly quenches their luminescence intensity and lifetime — severely reducing the conversion efficiency and output power of fiber lasers. One the other hand, the reaction between OH^- and phosphate would increase the amount of non-bridge oxygen in glass and destroy the network structure, thus, deteriorating the structure-related properties and impacting the long-term stability of optical fibers.

As shown in Fig. 1(a), we fabricated the Nd³⁺-doped aluminate, germanate, silicate, borate and phosphate glasses in air and measured their FTIR spectra for comparison. The broad asymmetrical absorption band near 3 µm belongs to the stretching vibration of free OH^-. As can be observed, the transmittance of Nd³⁺-doped phosphate glass is much lower than other oxide glasses and its infrared absorption edge even cuts off at 3 µm owing to the high absorption of OH^-. The absorption coefficient (${\alpha _{O{H^ - }}}$) of OH^- could be evaluated by the following equation [35]:

 

Fig. 1. (a) FTIR spectra of Nd³⁺-doped aluminate, germanate, silicate, borate and phosphate glasses fabricated in air. (b-d) FTIR spectra, luminescence spectra and decay curves of Nd³⁺-doped phosphate glasses fabricated in air atmosphere, with dry O₂ bubbling and with CCl₄ bubbling, respectively.

Download Full Size | PDF

Table 1. OH^- absorption coefficient (${\alpha _{O{H^ - }}}$) of Nd³⁺-doped phosphate, silicate, germanate, borate and aluminate glasses prepared in air

View Table

Introducing reactive dehydration agent in glass melt has been generally recognized as an effective technique for removing OH^- from laser glasses. Here, considering the excellent dehydration effect of carbon tetrachloride (CCl₄) that can eliminate OH^- by reaction of 4M-OH + CCl₄↔2M-O-M + CO₂↑ + 4HCl↑ (M: P⁴⁺) without introducing any new components into glass [35], it was bubbled into the Nd³⁺-phosphate glass melt by using dry O₂, during melting process. In order to compare dehydration effect, another Nd³⁺-phosphate glass was prepared by same procedure with only dry O₂ bubbling. The measured results of FTIR spectra are presented in Fig. 1(b). Notably, an obvious enhancement of the transmittance near 3 µm can be observed after the bubbling by dry O₂ and the introduction of CCl₄. Especially, the OH^- absorption band almost disappears and absorption coefficient is effectively decreased from 21.9 cm⁻¹ to 1.13 cm⁻¹ when the Nd³⁺-doped phosphate glass is dehydrated by CCl₄. To quantify the variation on radiative transition properties, we measured the luminescence spectra and decay curves of the glass samples under the excitation at 808 nm. As shown in Fig. 1(c-d), both the NIR luminescence intensity of Nd³⁺ and lifetime of ⁴F_3/2 level are remarkably improved by about 3 times, especially the lifetime is prolonged from 114 µs to 340 µs, which indicates that the energy transfer between Nd³⁺ and OH^- is largely cut off.

3.2 Fiber design and drawing

To draw high-quality optical glass fibers by Rod-in-Tube method and avoid irregular core-clad interface or stress-related fiber loss, the cladding phosphate glass composition was deliberately designed for the well-matched thermal properties with core glass. The refractive index difference between core and cladding glasses is 0.05, giving rise to a numerical aperture (NA) of 0.12. The normalized frequency (V) of step-index single-mode fiber that supports only fundamental mode (LP₀₁) must meets the following condition [36]:

Here, for single-mode operation at 0.9 µm, 2a is designed to be 5 µm and the V value at 900 nm is determined to be 2.09. Since the fiber diameter is the proportional reduction of the preform in Rod-in-Tube method, the cladding phosphate glass tube was fabricated and machined with an external diameter of 40 mm and an inner diameter of 1.6 mm. The core rod, on the other hand, was firstly reshaped into a cylindrical rod with a diameter of 10 mm and then drawn to the thin rod with a diameter of 1.6 mm, due to the difficulty of direct machining. Figure 2(a-b) presents the obtained core rods and the assembled fiber preform, which exhibit excellent optical quality. As shown in Fig. 2(c), the Nd³⁺-doped phosphate glass fiber was successfully drawn by precisely adjusting fiber drawing speed and the feed speed of the preform. Both the optical image (Fig. 2(d)) and back scattered electron image (Fig. 2(e)) of fiber cross section show that the optical fiber has a clear core-cladding structure. The core and cladding diameters are 5 µm and 125 µm respectively, conforming to above single-mode fiber design and initial core-cladding ratio of the preform. The elemental distribution of P, Ba, La and Nd characterized by EPMA are shown in Fig. 2(f-i). Despite the small intensity difference because of the close composition of core and cladding, the clear boundary between them can be clearly identified.

 

Fig. 2. (a) Photographs of the core rods with diameter of 10 mm (top) and 1.6 mm (low). (b, c) Photographs of the fiber preform and fabricated fiber. (d-e) Optical image and back scattered electron image of fiber cross sections. (f-i) EPMA images of the elemental distribution of P, Ba, La and Nd.

Download Full Size | PDF

The small signal gain at 915 nm was measured by using a 4.5-cm developed Nd³⁺-doped phosphate fiber. A LD at 915 nm was employed as signal source with an input power of 0.23 mW. The experimental setup and measurement procedure is similar to that of our previous studies [37]. Figure 3(a) illustrates the measured net gain as a function of the pump power provided by a LD at 808 nm. The net gain is first linearly increase with pump power and then tends to saturate. The measured saturable net gain is 12.15 dB, indicating a net gain coefficient of up to 2.7 dB cm⁻¹, which is much higher than that of the commercial Nd³⁺-doped silica fiber (only 1.0 dB cm⁻¹) [38]. Figure 3(b) displays the transmission loss at 1310 nm measured by cut-back method, and the fitting result shows that loss of Nd³⁺-doped phosphate fiber is 8.5 dB m⁻¹, which is relatively high and can be mainly attributed to the following three reasons. First, the transition metal impurities, such as Fe³⁺, are introduced in the glass melting process or even from raw materials, resulting in strong impurity absorption. Second, the loss stems from OH^-. Although we have significantly decreased absorption of OH^- by CCl₄, there is still a small amount of OH^- remaining in the glass. Third, the inner surface of cladding as well as the surface of core rod are not ideally polished, causing certain defects in the fiber core-cladding interface. It is believed that the fiber loss could potentially be further reduced by optimizing these three aspects.

 

Fig. 3. (a) Net gain at 915 nm versus pump power, measured for a 4.5-cm Nd³⁺-doped phosphate fiber. P_s: input signal power. (b) Fiber transmission loss at 1310 nm measured by the cut-back method.

Download Full Size | PDF

3.3 Generation of high-repetition-rate mode-locked pulses at 0.9 µm

As depicted in Fig. 4(a), we first constructed a CW fiber laser at 915 nm to explore the fiber performance. The laser cavity consists of a 4.5-cm-long Nd³⁺-doped phosphate glass fiber, a high reflection fiber Bragg grating (HR-FBG, >99.9%) with a 3-dB bandwidth of 0.4 nm and a low reflection fiber Bragg grating (LR-FBG, 70%) with a 3-dB bandwidth of 0.06 nm. The laser cavity is placed in a copper tube, which is temperature-controlled by a cooling system. A single-mode LD at 808 nm is fusion spliced to the pump port of 808/915 nm wavelength division multiplexer (WDM) and coupled into the cavity in a forward pump scheme. As shown in Fig. 4(b), the CW laser at 915 nm (red line) with a signal-to-noise (SNR) of 57 dB could be clearly observed. Inset in Fig. 4(b) is the closeup of the spectrum, showing the exact laser wavelength is 914.5 nm. In contrast, when the gain fiber is replaced by commercial Nd³⁺-doped silica fiber (Nufern, PM-NDF-5/125) with same 4.5-cm length, the laser SNR presents an obvious decrease (blue line). The output power was measured as a function of absorbed (red) and launched pump power (green), as shown in Fig. 4(c). It is noted that the output power linearly increases with pump power, yielding a slope efficiency of 11.2% and a threshold of 25 mW when taking Nd³⁺-doped phosphate fiber as gain medium. Yet, for the Nd³⁺-doped silica fiber, the slope efficiency is reduced prominently to 1.9%, while the threshold is increased to 75 mW.

 

Fig. 4. (a) Experimental setup of CW fiber laser at 915 nm. (b) Laser spectra of Nd³⁺-doped phosphate single-mode glass fiber (red line) and Nd³⁺-doped silica fiber (blue line). Inset show the closeup of the optical spectrum. (c) Output power of the 915 nm CW lasers that were constructed by the Nd³⁺-doped phosphate glass fiber and Nd³⁺-doped silica fiber, respectively. Red line is the output power versus absorbed pump power and green is versus launched pump power by Nd³⁺-doped phosphate glass fiber. Blue line is the output power versus absorbed pump power by Nd³⁺-doped silica fiber

Download Full Size | PDF

To demonstrate the uniqueness of the new Nd³⁺-doped phosphate fiber, we employed it to build an ultrashort laser cavity at 0.9 µm for mode-locking that delivers ultrashort pulse train with HRR. As presented in Fig. 5(a), we established an all-fiber passively mode-locked laser cavity composed of a semiconductor saturable absorber mirror (SESAM, 22% modulation depth and 1 ps recovery time), a fiber type dielectric film (DF, SiO₂/Ta₂O₅) and an 8.4 cm long Nd³⁺-doped phosphate fiber (NPF). The NPF was fixed inside a ceramic ferrule with an inner diameter of 125 µm by epoxy, and then both end facets were perpendicularly polished. The fiber type DF was butt-coupled to one end facet of NPF, then the fiber pigtail was fusion spliced to the common port of 808/915 nm WDM. The SESAM was connected with the other end facet of the NPF. The pump laser at 808 nm was coupled by fused splicing with the pump port of 808/915 nm WDM. Figure 5(b) depicts the average output power as a function of the launched pump power. The self-started CW mode-locking of the oscillator is achieved when the pump power is higher than 210 mW. The output power increases linearly with pump power and with a slightly improvement when entering the CW mode-locking. The optical spectrum of the output pulse extends from 898 to 904 nm with a peak wavelength of 899.5 nm (Fig. 5(c)). The autocorrelation trace of the oscillators is shown in Fig. 5(d). The FWHM of autocorrelation trace is 4.3 ps. Assuming a sech²-pulse shape and the additional stretching factor (1.54) caused by the convolution, the pulse duration is determined to be 3.2 ps. Figure 5(e) shows the mode-locked pulse train in a time span of 20 ns, which was received by a PD with a bandwidth of 12.5 GHz and then recorded by a real-time oscilloscope with a bandwidth of 20 GHz. Given the high net gain of the Nd³⁺- doped phosphate fiber, mode-locked pulses with a period of 840 ps is obtained from a laser cavity at a length of centimeter level, corresponding to a fundamental repetition rate of 1.2 GHz, the highest value so far — two orders of magnitude higher than previous works [24,27–29]. The repetition rate of the mode-locked pulses was characterized by the phase noise analyzer, as presented in Fig. 5(f), the fundamental frequency is located at 1.19 GHz, which is consistent with the period of the pulse train, and the SNR of the RF spectrum is up to 66.9 dB. The sidebands on both sides of the fundamental frequency are resulted from the weak modulation, which is caused by the birefringence effect on mode-locked pulse train [39]. The above results reveal that this Nd³⁺-doped phosphate glass fiber with high net gain is a promising gain medium for HRR UFLs at 0.9 µm — a perfect wavelength window for two-photon biophotonics.

 

Fig. 5. (a) Experimental setup of the high-repetition-rate mode-locked fiber laser at 0.9 µm based on Nd³⁺-doped phosphate fiber. (b) Output power of the mode-locked fiber laser as a function of pump power. ML: mode-locking. (c) Optical spectrum of the mode-locked fiber laser. (d) Autocorrelation trace of the mode-locked pulses. (e) Typical pulse train in 20 ns time span. (f) Radiofrequency (RF) spectrum of the mode-locked pulses. Inset is the RF spectrum in a broader frequency range.

Download Full Size | PDF

4. Conclusions

In summary, we have developed a Nd³⁺-doped phosphate glass fiber with sufficiently high gain at 0.9 µm. After effective removing of OH^- by the incorporation of CCl₄ during glass melting process, the glass OH^- absorption coefficient is reduced from 21.9 cm⁻¹ to 1.13 cm⁻¹. The Nd³⁺-doped phosphate fiber has an NA of 0.12, a core diameter of 5 µm, which enables single-mode operation at 0.9 µm. Both the gain coefficient and laser slope efficiency at 915 nm are greatly enhanced compared with that of commercial Nd³⁺-doped silica fiber. By utilizing passively mode-locking technique, an ultrafast fiber laser at 0.9 µm with a pulse repetition rate of 1.2 GHz is realized by using a short piece of this Nd³⁺-doped phosphate fiber, which is the highest repetition rate for mode-locked fiber laser at 0.9 µm as far as we know. These results indicate that this Nd³⁺-doped phosphate fiber is a promising gain fiber for HRR UFLs at 0.9 µm, which is anticipated to be promising for two-photon biophotonics.