Optical frequency combs linking optical and microwave frequencies are created using dissipative Kerr solitons (DKSs) in high-Q optical microresonators^[1-6]. Often referred to as “soliton microcombs,” these devices have already been widely used in various system-level information and metrology applications^[6-11]. These applications include radio frequency (RF) photonic filters^[12], optical frequency synthesis^[13,14], photonic neural networks^[15], optical quantum information^[16-19], coherent telecommunication^[20-23], and optical atomic clocks^[24].

Notwithstanding these developments, the deterministic access and manipulation of soliton states remain challenges that hinder the widespread application of soliton microcombs. Photothermal effects, which result from the absorption of light and thermo-optic effects, are the primary cause of this issue in optical microresonators, particularly those constructed on integrated platforms^[25]. A self-organized pulse waveform, or a soliton state, forms when a continuous-wave pump with a microresonator resonance scans its frequency from the blue resonant edge to the red resonant edge^[26]. However, the photothermal action causes strong thermal instabilities, which quickly leads to the destruction of the soliton states. Consequently, various complex methods have been devised to address this concern^[27]. The most widely utilized techniques for creating soliton combs include rapid pumping laser frequency scanning, power kicking^[1,26,28,29], self-injection locking scheme^[30,31], pulse pumping^[32], and auxiliary laser thermal compensation^[33-36]. Among these, rapid pumping laser frequency scanning and auxiliary laser thermal compensation are uniquely superior in program control, while the self-injection locking scheme and pulse pumping are temporarily difficult to program control due to the complexity of the algorithm and the more complicated mechanism of generating the optical frequency comb. Each of these schemes has unique qualities, and the one chosen to create a Kerr optical frequency comb can be based on the specific requirements and experimental circumstances of the experiment.

Although there have been numerous laboratory research on DKS combs, it still takes a considerable amount of time to improve them for industrial applications. While progress has been made, the widespread application of optical frequency combs faces significant challenges, which can be generated through either self-starting or program-controlled methods. These challenges must be overcome to make the technology more accessible to a broader range of users^[37-40]. Nowadays, many teams have worked on writing programs for controlling the generation of DKS combs. A computer-controlled dual-laser DKS comb system still stably accesses the single-soliton state even if the resonant frequency drifts by a few GHz, or the pump power drifts by 20%^[41]. Besides, based on a closed-loop control system, the repetition rate stability of the soliton microcomb source improved more than 20 times^[42]. Further, a program is developed to automatically address to target N-soliton state with a near 100% success rate and as short as 10 s time consumption^[43]. Therefore, the automatic and fast generation of a “one-button start” DKS source is highly desirable to transfer the development of DKS technology from laboratory proof-of-concept experiments to industrial applications.

In this work, the piezoelectric frequency scanning method with appropriate scanning speed and accuracy settings was experimentally utilized. A compact, reliable, and consistent start-up soliton optical frequency comb experimental system is constructed. In addition, a set of control programs using LabVIEW has also been written to realize the “one-button start” of the single soliton. The system with the matching software control program has two obvious main advantages, namely, the system is very stable and reliable after sealing, and the time to generate the single-soliton state optical frequency comb and the chaotic state optical frequency comb is very short, which is faster than the time to generate the single soliton in existing literatures. The “one-button start” makes it possible to produce a single-soliton state in an average of 1.5 s. A system stability test was conducted and achieved a 100% success rate by generating 1000 single-soliton states for a duration of >48h and 100 chaotic states.

The laser tuning technique has proven to be an effective method for creating DKS^[1]. In this process, the continuous wave (CW) pump laser is adjusted in a range from the blue-detuned to the red-detuned region^[44,45], which is referred to as “forward tuning.” Initially, the CW pump is tuned into the resonance of the microresonator and is in the blue-detuned regime. The slow thermal nonlinearity and fast Kerr nonlinearity of the microresonator cause the cavity resonance to shift and be locked to the pump laser over time^[46,47]. In this regime, the Kerr comb formation can be observed if the optical power in the cavity reaches the threshold of modulation instability (MI). This causes a triangular pattern in the generated comb light, over the pump frequency detuning. As the pump is further adjusted, it will transition into a red-detuned regime where multiple solitons can form simultaneously, leading to a multiple-soliton state. When the soliton state is achieved, a distinctive power trace appears in the comb light generation, characterized by a steep step. The height of the step corresponds to the number of solitons present in the state. Then, the “backward tuning” provides a way to reliably access the single-soliton state starting from an arbitrary multiple-soliton state^[26,48].

In the state of DKSs, the key parameter is the effective laser-resonance frequency detuning that determines both the intensity and the duration of soliton pulses. This detuning is defined as 2πδeff=ω0˜−ωp,(1)where ω0˜ indicates the frequency of the pumped cavity resonance and ωp is the pump laser frequency. Alternatively, the absolute detuning is defined as 2πδ=ω0−ωp,(2)where ω0 indicates the initial cold cavity resonance frequency. It has been shown that solitons are supported within a certain range of effective detuning when the pump is effectively red-detuned (ωp<ω0˜) with a constant power, which is referred to as the soliton existence range.

In addition, strong thermal effects can make the creation of soliton optical frequency combs extremely challenging. The precise explanation is that the energy in the cavity rapidly decreases when the pump laser scans from the blue-detuned point to the red-detuned point, which also causes a rapid reduction in the temperature of the microcavity. The detuning of the pump laser with respect to the cavity resonance will then abruptly increase, exceeding the maximum permitted detuning range of the existence of solitons, and it will not be able to realize the experimental generation of the soliton optical frequency comb. This is due to the thermo-optic effect, which causes the microcavity resonance mode frequency to be blue-shifted while the pump laser is correspondingly red-shifted. The experimental investigation will employ the piezoelectric frequency rapid scanning approach to minimize the impact of heat effects in the microcavity and ultimately achieve the successful creation of the soliton optical frequency comb.

One of the key limiting factors of soliton microcombs is that their operation does not solely rely on the properties of the microresonators used for the generation of DKSs. While the microresonator parameters, such as Q-factor, dispersion profile, bus-waveguide coupling, and others, indeed play a major role in defining the resulting spectrum, access to soliton states, and their dynamics, the microresonator itself is a passive device, which unavoidably needs a driving laser for continuous operation. Furthermore, launching DKS states can be rather challenging and require complex tuning mechanisms. A few steps toward such systems have already been made recently with chip-scale laser sources. However, the system operation either was not stable or required additional equipment and a lot of fine-tuning and adjustments for the microcomb operation. As a result, all microcomb applications so far have been mostly limited to the laboratory environment with constant monitoring of the system operation, short operation time scales, and usage of unique costly, large-scale equipment that invalidated the size, weight, and power (SWaP) advantages of the soliton microcomb technology. On the other hand, to continue leveraging the SWaP advantages of microcombs, a system needs to be developed that not only possesses the required engineering challenges but also employs intelligent software capable of initiating, stabilizing, and controlling the system with minimal user intervention. The primary goal of the hardware work was to design a compact system architecture with a limited set of components to ensure stable operation of the soliton state in chip-integrated devices after the launch of DKS states. This would enable reliable and repeatable launching of DKS states and guarantee the stable operation of the soliton state. Therefore, a soliton optical frequency comb generation system is built, which is shown in Fig. 1.

First, the free spectral range (FSR) of the Si3N4 microresonator features 100.3 GHz and an intrinsic quality factor of Q0≈1.7×107, which is packaged in a black box. A 450mm×500mm chassis of 2U height (1U equals 4.445 cm) was used, which is supposed to facilitate the mounting of the system components inside^[49]. Second, a smaller, more compact version of a seed laser (NKT laser) with a piezo voltage amplifier operating at 1550 nm, a compact erbium-doped fiber amplifier (OEM EDFA with a maximum output power of 500 mW), and a multifunction I/O device (NI USB 6003, 100 kS/s) serving as both a data acquisition system and a driving function generator for the pump laser were used in the system. Third, a temperature stabilization subsystem was added to the system. It permits the adjustment of optical power-induced chip heating or maintains the stability of the chip temperature against variations in the surrounding air temperature with an accuracy better than 0.01°C. The ability to fine-tune the FSR or the absolute resonance places is made possible by the chip’s temperature control. Fourth, an internal switching power supply with an output of 12 V was installed in the system to provide power to all active components, which are distributed from a single standard 220 V power plug. Ultimately, all of the system’s controls are routed through a single USB wire that is connected to an external computer by means of a small USB hub.

The control software for the soliton optical frequency comb generation system was developed based on LabVIEW. The operation interface of the whole program is divided into three parts, namely, the photodiode (PD) power monitoring part, the “one-button start” parameter adjustment part, and the feedback control part. The most important part of this is the “one-button start” part, which has logic codes and an algorithm flowchart, as shown in Fig. 2.

The program logic is divided into two parts, where the red part aims at fast frequency sweeping to a multi-soliton state and the blue part aims at backward tuning from a multi-soliton state to a single-soliton state. We divide the entire power trajectory for generating the single-soliton state into three parts in terms of power magnitude, which are zone I, zone II, and zone III. (The classification is based on the magnitude of the optical frequency comb power of different soliton states.) We have normalized the voltage values obtained from the PD by multiplying the received PD voltages by the corresponding weights in LabVIEW through extensive experiments so that the PD voltage values of the single-soliton states become values between 1 and 2, laying the foundation for the “one-button start” process afterward. When the “one-button start” is touched, the whole program starts running. First, set the maximum voltage value of the sweep to X, the minimum voltage value to 0, and the sweep time to 0.02 s. If the PD voltage P lands in zone I, meaning that the laser frequency is blue-detuned and the state is chaotic comb, then the output voltage X should be set to X+0.1. If the PD voltage P lands in zone III, meaning that the soliton state is not yet reached and the laser frequency scanning window needs adjustment, then the output voltage X should be set to X−0.1. The process continues until the PD voltage lands in zone II, representing that the multi-soliton state is reached, and the red-colored part shown in Fig. 2 ends. Next, the output voltage X begins to decline slowly until the PD voltage P lands in zone III, signaling that the multi-soliton state progresses to the single-soliton state. In case the PD voltage P does not reach zone III, the system reinitiates and repeats the process.

Correspondingly, we further developed the feedback adjustment part, as shown in Fig. 3(a), which can do the stable adjustment of the generated single-soliton state optical frequency comb, based on the principle of monitoring the PD voltage and setting the upper and lower limits of the PD voltage as well as the step voltage. If the PD voltage exceeds the set value, the output voltage is adjusted in steps of the adjustment voltage. If the PD voltage value is higher than the upper limit, the output voltage has to be subtracted from the step voltage. If the PD voltage value is lower than the upper limit, the output voltage is added to the step voltage. These continue until the PD voltage value is within the set interval. This feedback adjustment method allows the generated single-soliton state to operate stably for more than 2 days. The system was connected to the spectrometer and triggered the feedback regulation module at the same time to keep the system continuously and stably generating the single-soliton state optical frequency comb, and then it extracted the data recorded by the spectrometer, the results of which are shown in Fig. 3(b).

To further demonstrate that our experiments produced the chaotic state, the multi-soliton state, the dual-soliton state, and the single-soliton state, the spectral images were extracted corresponding to three zones throughout the experiment, as shown in Fig. 4.

As we can clearly see from Fig. 4, the accepted voltage power range was divided into three zones using PD voltage as a criterion, and the spectrum corresponding to each regional junction corresponds exactly to different soliton states. As shown in Fig. 4(b), zone I corresponds to the chaotic state shown in Fig. 4(c)(i). The spectrogram corresponding to the junction of zone I and zone II is shown in Fig. 4(c)(ii). This is a multi-soliton state optical frequency comb. The spectrogram corresponding to the junction of zone II and zone III is shown in Fig. 4(c)(iii). This is a dual-soliton state optical frequency comb. The spectrogram of the single-soliton state optical frequency comb is shown in Fig. 4(c)(iv).

Following this, the overall stability of the software and hardware was tested. The “one-button start” was triggered 1000 times consecutively, generating 1000 single-soliton optical frequency combs, as shown in Fig. 5(a). Moreover, 100 chaotic states were generated consecutively in 31 s, as shown in Fig. 5(c). In order to show the power trajectory of the PD voltage more clearly, the production of the single-soliton state was also tested 10 times in the 1000 to 1015.6 s time interval, as shown in Fig. 5(b), and the chaotic state 10 times in the 0 to 4 s time interval, as shown in Fig. 5(d). Furthermore, the time distribution of 1000 single-soliton state production was tabulated and obtained in Fig. 5(e).

From Fig. 5(a), it takes 1517.0 s to produce 1000 single-soliton states, and the average time required for each generation is 1.51 s. By analyzing the PD voltage, single-soliton states arrive at a 100% success rate for these 1000 times. The results of the entire 1000 generated experiments were statistically analyzed in histograms. From the statistical histogram, it can be clearly seen that most of the startup time stays between 1.1 and 2.1 s, of which 56.7% is within 1.5 s, 39.1% is between 1.6 and 2 s, and 4.2% is more than 2 s. Based on these statistics, it can be concluded that the control program has been designed to be stable and reliable, and it can control the time to produce a single soliton within 2 s. Moreover, chaotic microcombs have emerged as promising sources for various applications such as private communication, encryption, anti-interference sensing, and reinforcement learning. It is also necessary to test the stability of generating the chaotic states. Therefore, 100 chaotic states were generated continuously and recorded a total time of 31 s, with an average of each time being 0.31 s.

Here, this work is compared with the work of other teams, as shown in Table 1. This work can compress the time to produce a single-soliton state to about 1.5 s, which is superior to the average 14.6 s result in Ref. [43]. At the same time, an optical comb generation system was designed in which the optical comb generation and monitoring units were mounted inside an aluminum box, leaving only a computer outside for condition monitoring, and we tested the stability of the system for generating a single-soliton state optical comb with up to 2 days of continuous generation. In addition to this, we have focused on the generation of chaotic optical frequency combs and have implemented program control of them.

In conclusion, the generation link of a single-soliton state optical comb was integrated and packaged in a 2U aluminum box, and a software package was developed to enable programmed control of the “one-button start” generation of the soliton state optical comb. To assess the performance and stability of the whole system, the “one-button start” was triggered 1000 times, obtaining 100% success in generating the single-soliton state, with a total time of 1517.0 s and an average time of 1.5 s, and the single-soliton state optical frequency comb can exist continuously beyond 2 days. In the process of debugging the program control, the stability of chaotic state optical frequency comb generation was also checked and was generated 100 times with a total time of 31 s. This work provides new progress for the programmed generation of chip-based optical frequency combs.