As the next-generation intelligent integrated digital infrastructure, 6G will intersect and integrate with information technologies such as artificial intelligence, big data, and advanced computing, achieving deep coupling of communication with perception, computation, and control. It possesses typical characteristics such as ubiquitous connectivity, inclusive intelligence, multi-dimensional perception, comprehensive coverage, green and low-carbon, and inherent security^[1,2]. Huawei company stated that 6G, like a massive distributed neural network, integrates communication, perception, and computing capabilities, deeply fusing the physical, biological, and digital worlds, truly ushering in a new era of “Everything Intelligent Connectivity”^[3,4].

Research indicates that potential application scenarios of 6G include immersive extended reality, holographic communication, sensory interconnection, smart interaction, communication perception, and digital twin^[5,6]. It can empower vertical industries such as the industrial Internet^[7–9], Internet of Vehicles^[10,11], smart cities^[12,13], smart grids^[14,15], smart agriculture^[16], and smart healthcare^[17,18], promoting the transition of humanity from the information age to the intelligent age, from the flat era to the three-dimensional era. The IMT-2030 (6G) Promotion Group defines the vision of 6G networks with two inherent characteristics (inherent intelligence and inherent security) and three integrations (integration of air, space, and ground; integration of communication and perception; and integration of computation and networking). In the future, the 6G transport network will be designed as an integrated computational and networking carrier, aiming to fully leverage the entrance advantages of the 6G network for users. It will use networking as the foundation, leveraging supercomputing/smart computing and capability resource pools to create differentiated computing services and realizing open service capabilities and collaborative operations of the integrated computing and networking system. This will establish a new type of computing and networking collaborative architecture with inherent computing power and inherent security^[1].

With the growth of ubiquitous intelligent services such as holographic interaction, the metaverse, and smart transportation, as well as the increasing demand for AI-driven computation, there has been a significant surge in bandwidth requirements at the infrastructure resource layer. Optical network bandwidth continues to expand, with the backbone transmission network entering the era of 400G and long-distance 800G technology on the horizon^[19,20]. Moreover, in the 5.5G and 6G eras, the rates of fronthaul, midhaul, and backhaul networks in mobile communication systems are also continuously increasing. However, with the improvement of optical communication rates, increased optical power, and extended transmission distances, the accumulation of nonlinear effects has caused significant damage to optical communication signal transmission, becoming one of the major factors restricting system performance. Hence, research on advanced optical modulation formats has been a key focus of researchers^[21–23]. The frequency shift keying (FSK) modulation format, due to its ability to carry different service signals through different frequencies with constant power and high nonlinear resistance, has gained widespread attention in recent years^[24–27]. However, existing research on single-channel systems mainly focuses on rates below 40 Gbps^[25–27]. This study proposes an advanced alternating polarization chirped return-to-zero frequency shift keying (Apol-CRZ-FSK) modulation format for 100 Gbps optical communication systems that can meet the integrated computing and networking carrying requirements of future 6G. Based on the FSK modulation format, the chirp amount is changed to offset the impact of dispersion and enhance the dispersion tolerance of the signal. In addition, the orthogonal polarization states of adjacent bit optical signals reduce their power interactions and improve the nonlinear tolerance.

In the Outlook on 6G Network Architecture released by the 6G Promotion Group at the global 6G Development Conference in 2023 for the first time, the 6G network architecture is divided into three layers, mainly including the infrastructure resource layer, network function layer, and application and openness layer. As shown in Fig. 1, the application and openness layer provides multidimensional capability opening for holographic interaction, industrial Internet, smart transportation, smart healthcare, the metaverse, and other application scenarios of the 6G era through services such as data, computing, perception, AI, connectivity, and security. The network function layer provides unified scheduling of multiple elements for the upper-layer applications and openness through connection, computation, AI, data, perception, and other functions. However, all these functionalities rely on the supply of various types of basic resources in the infrastructure resource layer, which mainly includes infrastructure such as computing power and networking resources. In the era of 6G, networking and computing are deeply integrated, with networking promoting computing and computing supporting networking. Networking and computing perceive each other, coordinate with each other, realize real-time and accurate computing resource discovery, realize flexible and dynamic service scheduling, provide ubiquitous computing and services, achieve rational allocation of computing resources and user unawareness, empower consistent user experiences, improve the efficiency of network and computing resource utilization, and realize efficient collaboration among cloud, edge, network, and computing.

In the infrastructure resource layer of the 6G integrated computing and networking architecture, the Apol-CRZ-FSK modulation format can be applied in 100 Gbps optical cross connect (OXC) optical transmitters and receivers. As shown in Fig. 1(a), the optical signal emitted by the laser CW1 of the Apol-CRZ-FSK signal transmitter is sent to the Mach–Zehnder modulator 1 (MZM1), which operates in a push–pull state with the amplitude of the RF driving voltage set to half the half-wave voltage, and the direct current (DC) bias is set at the zero-crossing point of the transmission curve. At this time, the carrier suppression can be achieved by modulating the MZM1 to obtain the required two wavelengths, and the wavelength interval is twice the frequency of the RF signal. The resulting signal is then sent to a phase modulator for differential phase shift keying (DPSK) modulation of the two different frequency optical signals, generating two DPSK signals that are subsequently transmitted to MZM2.

$V_{\pi }$ represents the half-wave voltage of the modulator, and the bias of MZM2 is set at $\frac{V_{\pi }}{2}$. The amplitude and frequency of the clock signal are $\frac{V_{\pi }}{4}$ and $B$, respectively, and the phase difference between the two clock signals is $\pi$, while the symbol $t$ denotes the time variable, customarily employed to delineate physical quantities that are time-varying, thereby facilitating the depiction of dynamic characteristics pertinent to communication systems. The voltages $V_1$ and $V_2$ on the two arms are given by$$V_1=\frac{V_{\pi }}{4}+\frac{V_{\pi }}{4}\cos (2\pi Bt),$$(1)$$V_2=-\frac{V_{\pi }}{4}+\frac{V_{\pi }}{4}\cos (2\pi Bt+\pi ),$$(2)$$V_1+V_2=0,$$(3)$$V_1-V_2=\frac{V_{\pi }}{2}+\frac{V_{\pi }}{2}\cos (2\pi Bt).$$(4)

Therefore, the optical field and optical intensity $I$ of the two frequency signals output by MZM2 can be expressed as follows:$$E_{1\text{out}}=E_{1\text{in}}\sin (\frac{\pi }{4}-\frac{\pi }{4}\cos (2\pi Bt)),$$(5)$$E_{2\text{out}}=E_{2\mathrm{in}}\sin (\frac{\pi }{4}-\frac{\pi }{4}\cos (2\pi Bt)),$$(6)$$I_{1\text{out}}=I_{1\text{in}}{\sin }^2(\frac{\pi }{4}-\frac{\pi }{4}\cos (2\pi Bt)),$$(7)$$I_{2\text{out}}=I_{2\text{in}}{\sin }^2(\frac{\pi }{4}-\frac{\pi }{4}\cos (2\pi Bt)).$$(8)Let $\frac{I_{\mathrm{out}}}{I_{\text{in}}}=\frac{1}{2}$, then, ${\sin }^2(\frac{\pi }{4}-\frac{\pi }{4}\cos (2\pi Bt))=\frac{1}{2}$, $\pi Bt=\frac{3\pi }{4}$ or $\frac{\pi }{4}$. Thus, the duty cycle is $\frac{\frac{3\pi }{4}-\frac{\pi }{4}}{\pi }=\frac{1}{2}$, which corresponds to a 50% duty cycle. This results in the generation of an RZ-DQPSK signal, where the return-to-zero (RZ) denotes that the signal returns to zero at the midpoint of each symbol period.

Following this, the signal processing by a 1-bit delay Mach–Zehnder delay interferometer (MZDI) generates a return-to-zero frequency shift keying signal carrying optical pulses at each bit interval, which is then transmitted to the phase modulator. The phase modulator is utilized to control the chirp, producing a CRZ-FSK signal that is subsequently sent to the polarization alternator. The CRZ-FSK signal is split into two paths by the polarization beam splitter (PBS), with each path containing orthogonally polarized light with equal intensity. One path of polarized light directly enters the polarization beam combiner (PBC), while the other path of polarized light passes through the second phase modulator to impose modulation signals with phases of 0 and $\pi$ before entering the PBC. The PBC combines the two paths of polarized light to obtain signal light with orthogonally polarized states for adjacent bits, constituting the Apol-CRZ-FSK optical signal. As shown in Fig. 1(b), an incoherent single-ended detection method can be directly employed on the receiver side for frequency discrimination, enabling signal demodulation reception using a direct detection receiver, which is simple, practical, and cost-effective.

This research employs VPItransmissionMaker Optical Systems version 9.3 for the comprehensive simulation of the transmission system, with the signal transmission rate set at 100 Gbps. On the transmission line of the high-speed optical communication system, an erbium-doped fiber amplifier (EDFA) is initially employed to amplify the signal power, facilitating its long-distance transmission. The dispersion compensation fiber can be placed before the single-mode fiber (SMF), known as pre-compensation, after the SMF as post-compensation, or evenly placed before and after the SMF as hybrid compensation. On the receiving side, the signal should be amplified by an EDFA to compensate for the optical power loss during long-distance transmission and then received and demodulated by a direct detection receiver. Table 1 displays the parameter settings for the fiber link.

This study compared the transmission performance differences of 100 Gbps Apol-CRZ-FSK, CRZ-FSK, and differential quadrature phase shift keying (DQPSK) signals with pre-, post-, and hybrid dispersion compensation.

The relationship for the Q-factor variations of three different modulation signals under the same conditions as the launch power ranged from $-5\mathrm{dBm}$ to 30 dBm, and the single span distance ranged from 30 km to 230 km, are illustrated in Figs. 2(a)–2(c), Figs. 3(a)–3(c), and Figs. 4(a)–4(c), respectively.

As the launch power gradually increases indicated in Figs. 2(a)–2(c), the Q-factor of the signal at the same transmission distance gradually increases, reaching its maximum value, and then it gradually decreases. When the launch power is $-5\mathrm{dBm}$, the Q-factors of the signals transmitted over 90 km under pre-, hybrid-, and post-compensation are around 6 dB, with no significant difference. However, when the launch power is 14 dBm, the Q-factor of the received signal increases to 25 dB under the post-compensation [shown in Fig. 2(c)], which is 1 dB and 5 dB higher than that under the hybrid-compensation [shown in Fig. 2(b)] and pre-compensation schemes [shown in Fig. 2(a)], respectively, indicating better transmission performance under post-compensation. As the launch power continues to increase, the influence of nonlinear effects gradually increases, resulting in a decrease in the Q-factor. When the launch power is 24 dBm, the Q-factor of the signal starts to be less than 6 dB, indicating severe degradation of the signal and the inability to demodulate the signals normally. When the launch power is 19 dBm, the longest transmission distance of a single span segment can be achieved under pre-compensation conditions, reaching 175 km as shown in Fig. 3, while the optimal performance of the system can be achieved when the launch power is 23 dBm.

Transmission performance of 100 Gbps CRZ-FSK signals under three compensation schemes is shown in Figs. 4(a)–4(c). Similar to the transmission curve of Apol-CRZ-FSK signals, the Q-factor of the three compensation schemes is around 6 dB under the condition of 90 km and the launch power of $-5\mathrm{dBm}$. As the launch power increases, the Q-factor also increases. When the launch power is 9 dBm for the pre-compensation [shown in Fig. 4(a)], the 100 Gbps CRZ-FSK signal achieves the best transmission performance at 90 km, with a Q-factor of 21 dB, which is 4 dB higher than that under the hybrid-compensation [shown in Fig. 4(b)] and post-compensation schemes [shown in Fig. 4(c)]. When the launch power is 14 dBm, the Q-factor is 17 dB, which is 8 dB lower than that of Apol-CRZ-FSK signals under the same conditions. When the launch power is 16 dBm, the 100 Gbps CRZ-FSK signals can achieve the longest transmission distance of 155 km, which is 20 km less than that of Apol-CRZ-FSK signals. Apol-CRZ-FSK signals exhibit better transmission performance under the same conditions.

Figures 5(a)–5(c) illustrate the transmission performance of 100 Gbps DQPSK signals. The Q-factor of the three compensation schemes is around 4 dB while the signal transmission distance is 90 km and the launch power is $-5\mathrm{dBm}$, which is 2 dB lower than that of CRZ-FSK signals and Apol-CRZ-FSK signals. When the launch power is 7 dBm, the best transmission performance at 90 km is achieved under the pre-compensation scheme [shown in Fig. 5(a)], with Q-factor of 16.7 dB, which is 3 dB and 0.3 dB higher than that of hybrid compensation [shown in Fig. 5(b)] and post-compensation [shown in Fig. 5(c)], respectively. But it is 2 dB lower than the Q-factor of CRZ-FSK signals and 4 dB lower than the Q-factor of Apol-CRZ-FSK signals. When the launch power is 14 dBm, the Q-factor is 6 dB, which is 19 dB and 11 dB lower than that of Apol-CRZ-FSK signals and CRZ-FSK signals, respectively. When the launch power is 13 dBm, the longest transmission distance of 134 km can be achieved, which is 41 km and 21 km less than that of Apol-CRZ-FSK and CRZ-FSK signals. Apol-CRZ-FSK signals also exhibit the best transmission performance.

Figure 6 compares and analyzes the relationship between the bit error rate (BER) and the launch power for three signals at 80 km. When the launch power is 13 dBm, the 100 Gbps Apol-CRZ-FSK signal exhibits the best BER performance, significantly better than that of the other two signals.

This study investigated the relationship between the launch power and the Q-factor of 100 Gbps Apol-CRZ-FSK, CRZ-FSK, and DQPSK signals at these typical span transmission distances. As shown in Fig. 7, when the launch power is $-5\mathrm{dBm}$, the differences in Q-factors of the three signals after transmission distances of 80 km, 100 km, and 120 km are minimal. The Apol-CRZ-FSK signal exhibits the best transmission performance after 80 km transmission, followed by CRZ-FSK and DQPSK signals. At this point, the launch power is relatively small, and the impact of nonlinear effects is not significant.

As the launch power gradually increases, the Q-factors of the three signals also increase gradually. After reaching a peak value, nonlinear effects become the main factor affecting system performance with further increases in power and distance. The Q-factor gradually decreases as the launch power increases and the distance lengthens, with larger nonlinear effects leading to greater differences in Q-factor among the three signals. Figure 7 shows that the Q-factor of Apol-CRZ-FSK signals is 7.5 dB and 9.6 dB higher than that of CRZ-FSK and DQPSK signals, respectively, while the best transmission performance conditions are achievable at 80 km. Under the best transmission performance conditions achievable at 100 km, the Q-factor of Apol-CRZ-FSK signals is 12.9 dB and 16.5 dB higher than that of CRZ-FSK and DQPSK signals, respectively. The Q-factor of Apol-CRZ-FSK signals is 10.2 dB and 13.6 dB higher than that of CRZ-FSK and DQPSK signals, respectively, at 120 km.

The performance of 100 Gbps Apol-CRZ-FSK signals after transmission over multiple span distances is analyzed in Fig. 8. Each span consists of a segment of SMF, a dispersion compensating fiber, and an EDFA used to compensate for the signal’s long-distance transmission power loss. The number of spans ranges from 5 to 100, and each span distance varies gradually from 30 km to 120 km. The launch power ranges from $-10\mathrm{dBm}$ to 10 dBm. A four-dimensional relationship graph showing the number of transmission span, distance per span, launch power, and Q-factor is shown in Fig. 8. The research indicates that, when the length of SMF for each span is 31 km, it is possible to achieve transmission of 100 Gbps Apol-CRZ-FSK signals over 85 spans, with the longest transmission distance being 2635 km. At this point, the best transmission performance is achieved with a launch power of 4 dBm. For transmission distances of 40 km and 50 km, the end-to-end transmission distances are 1240 km and 850 km, respectively.

When the SMF length of each span is 60 km, the optimal transmission performance is achieved with an launch power of 7 dBm, resulting in a Q-factor of 9 dB for 600 km transmission; for signal transmission over 960 km, the optimal transmission performance is achieved with a launch power of 9 dBm.

This study proposes an advanced modulation format Apol-CRZ-FSK, which meets the requirements of integrated computing and networking for future 6G. Based on the FSK modulation format, the chirp amount is changed to offset the impact of dispersion and enhance the dispersion tolerance of the signal. In addition, the orthogonal polarization states of adjacent bit optical signals reduce their power interactions and improve the nonlinear tolerance. An incoherent single-ended detection method can be directly employed on the receiver side for frequency discrimination, enabling signal demodulation reception using a direct detection receiver, which is simple, practical, and cost-effective. Furthermore, comprehensive comparison and analysis of the transmission performance of 100 Gbps Apol-CRZ-FSK, CRZ-FSK, and DQPSK signals are conducted under identical conditions. The research indicates the high nonlinearity resistance capability exhibited by Apol-CRZ-FSK, highlighting its superior transmission performance.