TS-IJMR consists of three constant-flow pumps (2 PB-20005II, Beijing Xingda Science and Technology Development Co. Ltd.), two micro T-junctions (SS-1UTF, Beijing Xiongchuan Technology Co. Ltd.), and several stainless-steel capillary tubes (inner diameter d _i = 0.6 mm). Additionally, three capillary tubes link to the first micro T-junction to construct an internal structure similar to a “T-shaped micromixer” for the homogeneous mixing of Ni²⁺/Co²⁺ and GO solutions. At the second micro T-junction, two capillary tubes are aligned coaxially and inversely. However, the second micro T-junction outlet is no longer connected to the capillary tube but forms an enlarged channel, which can effectively reduce the blockage problem during precipitation. After the uniformly mixed Ni²⁺/Co²⁺/GO solution collides with reducer within the second micro T-junction at high velocities, the generated precipitates will immediately downward into other vessels for aging. Precipitation and aging processes are hence separated in two containers in the TS-IJMR route, which can provide a more uniform and steady environment for both particle nucleation and crystal growth on top of the enhanced micromixing performance and better process control of TS-IJMR. The geometric structures of TS-IJMR are depicted in Supplementary Figure S1.

Nickel sulfate hexahydrate (NiSO₄ · 6H₂O), Cobalt nitrate hexahydrate (Co(NO₃)₂ · 6H₂O) and potassium hydroxide (KOH) were provided by Aladdin Reagent (Shanghai) Co., Ltd. Potassium borohydride (KBH₄) was purchased from Shanghai Macklin Biochemical Co., Ltd. All the chemicals were analytical grade and used without further purification.

The controllable and scale-up synthesis of NCBO/RGO composites in TS-IJMR is shown in Figure 1. Firstly, 0.1 mol L⁻¹ NiSO₄ and Co(NO₃)₂ mixed solution (Ni²⁺: Co²⁺ = 3:1) and 1 mg ml⁻¹ GO solution were pumped into the first micro T-junction at constant volumetric flow rates to obtain a uniformly mixed Ni²⁺/Co²⁺/GO solution. Then, the mixture immediately flowed into the second micro T-junction to produce NCBO/GO-3 (i.e., Ni²⁺/Co²⁺ molar ratios of 3:1) precursor by reversely high-speed impact with 0.1 mol L⁻¹ KBH₄ solution. The volumetric flow rates of Ni²⁺/Co²⁺ solution, GO solution, and KBH₄ solution were maintained at 80, 80, and 160 ml min⁻¹, respectively, with the corresponding flow velocities of 4.76, 4.76, and 9.52 m s⁻¹, respectively. The precipitate generated from the second micro T-junction immediately flowed into a storage tank and aged for 6 h under room temperature with vigorous agitation. Next, the NCBO/GO-3 precursor was washed 3 times each by deionized water and anhydrous ethanol before being dried overnight at 80°C. Finally, the NCBO/GO-3 precursor was calcined at 300°C in an argon atmosphere for 3 h. The GO was successfully converted to reduced graphene oxide (RGO), thus producing NCBO/RGO-3. In addition, the samples synthesized in the same TS-IJMR route with Ni²⁺/Co²⁺ molar ratios of 1:2, 1:1, 2:1, and 4:1 were denoted as NCBO/RGO-0.5, NCBO/RGO-1, NCBO/RGO-2 and NCBO/RGO-4, respectively.

With the same synthesis method, nickel boride@borate/RGO (NBO/RGO) and nickel boride@borate/RGO (CBO/RGO) were synthesized by replacing the Ni-Co mixed solution with 0.1 mol ml⁻¹ NiSO₄ and 0.1 mol ml⁻¹ Co(NO₃)₂ solution, respectively.

The NCBO/RGO-S was synthesized by parallel precipitation in STR. First, 0.1 mol L⁻¹ NiSO₄/Co(NO₃)₂ mixture solution (Ni²⁺: Co²⁺ = 3:1) and 1 mg ml⁻¹ GO solution were added dropwisely at 10 ml min⁻¹ to a beaker with vigorous stirring to obtain the mixed Ni²⁺/Co²⁺/GO solution (Supplementary Figure S1). Then, Ni²⁺/Co²⁺/GO mixture solution and 0.1 mol L⁻¹ KBH₄ solution were added dropwise at 10 ml min⁻¹ to a beaker with vigorous stirring. After titration, the suspension was further aged for 6 h under room temperature with vigorous agitation. The subsequent steps are consistent with those of TS-IJMR. The resulting sample was labeled “NCBO/RGO-S”.

The crystal structure and phase composition of the TMBO/RGO samples were characterized using a X-ray diffraction analyzer (XRD, Bruker D8) with Cu Kα radiation (λ = 0.15406 nm) and X-ray photoelectron spectra (XPS, ESCALAB 250) with an Al Kα radiation source. Fourier transform infrared spectrum (FT-IR) was conducted in the wavelength of 4000–400 cm⁻¹ with a Bruker Vetex 70 spectrometer. The surface morphologies of the TMBO/RGO samples were observed with scanning electron microscopy (SEM, Nova NanoSEM 200) and transmission electron microscopy (TEM, JEOL JEM-2100). High-resolution transmission electron microscopy (HR-TEM), energy-dispersive X-ray spectrometer (EDS) and selected area electron diffraction (SAED) were performed on the same transmission electron microscopy. The nitrogen adsorption–desorption isotherms of the TMBO/RGO samples were recorded by a Micromeritics analyzer (ASAP 2020M), while their corresponding specific surface areas were determined from the Brunauer–Emmett–Teller (BET) equation.

The morphologies of TMBO/RGO samples were observed with electron microscopy. As shown by SEM images in Supplementary Figure S3, all the three TMBO/RGO samples consist of TMBO nanoparticles that tightly wrapped with RGO flakes. TEM images of the three TMBO/RGO samples are shown in Figures 2A–C. With respect to NBO/RGO, core-shell spherical NCO particles with a particle size of 30–40 nm and a shell layer thickness of 2–3 nm are loaded on RGO flakes (Figure 2A), while CBO/RGO contains many long strips with lengths of several hundred nanometers that are randomly dispersed on RGO flakes (Figure 2B). In addition, it can be seen from Figure 2C that the NCBO/RGO-3 possesses nanoflake structure without obvious agglomeration and accumulation. The corresponding HR-TEM image shows faint lattice fringes of 0.223 nm and interplanar spacing of 0.2 nm, which can be attributed to the (211) crystal plane of Ni₃(BO₃)₂ and the (102) crystal planes of Ni₃B, respectively. The low-resolution lattice fringes suggest that NCBO/RGO-3 has a low crystallinity (Figure 2D). The presence of (303) crystal plane of Ni₃(BO₃)₂, (102) and (211) crystal planes of Ni₃B in NCBO/RGO-3 is further confirmed by the SAED in Figure 2E. In addition, Figure 2F reveals the presence of agglomerated lamellar structure in the NCBO/RGO-S synthesized by STR route. Finally, the EDS mapping demonstrates a uniform distribution of Ni, Co, O, B, and C elements within the NCBO/RGO-3 (Figure 2G–K).

The N₂ adsorption-desorption isotherms are shown in Figure 3A to study the textural properties and pore structures of the three TMBO/RGO samples. All samples show the typical IV adsorption behaviors with distinct hysteresis loops in the relative pressure (P/P₀) range of 0.4–1.0, indicating they mainly comprises mesoporous structures. The pore size distributions derived from their corresponding desorption curves are shown in Figure 3B. All the samples display a sharp and narrow peak at 2–5 nm that presumably stem from the internal pore sizes of each TMBO/RGO. In addition, the pore size distribution peaks of 10–50 nm existed for NBO/RGO and CBO/RGO may be originated from the interstices between the loaded particles on the RGO sheet. These abundant pores, especially the 2–5 nm mesopores, can increase the specific surface area of the material (Borchardt et al., 2018). Therefore, the BET specific surface area of NCBO/RGO-3 is as high as 62.37 m²/g, which is much larger than 30.31 and 42.89 m²/g of NBO/RGO and CBO/RGO. A larger specific surface area provides more superficial electroactive sites to participate in redox reactions, while abundant pores can shorten the transport channels from electrolyte ions to the internal pores of electrode materials (Borchardt et al., 2018). Therefore, NCBO/RGO-3 with relatively large BET specific surface area and abundant pore channels is expected to display high electrochemical performance.

The XRD patterns of the three TMBO/RGO samples are shown in Figure 4A. No obvious diffraction peaks are observed suggests the amorphous structure of CBO/RGO. The broad diffraction peaks of NBO/RGO at 34.2° and 61.3° can be attributed to the (202) and (303) crystal planes of Ni₃(BO₃)₂ (PDF#26-1284), respectively, while the sharp diffraction peak at 44.7° corresponds to the (102) crystal plane of Ni₃B (PDF#48-1223) (Chen et al., 2019a). In addition, NCBO/RGO-3 also shows a Ni₃B diffraction peak at 44.7°, but its peak intensity is significantly weaker than that of NBO/RGO, indicating that the doping of Co elements significantly reduces the crystallinity of NCBO/RGO-3.

The FT-IR spectrum of NCBO/RGO-3 is shown in Figure 4B. The two peaks located at 3440 and 1631 cm⁻¹ can be ascribed to the O-H stretching and bending vibrations of absorbed water within the material skeleton, respectively. It is found that two characteristic peaks appear at 1278 and 1423 cm⁻¹ in NCBO/RGO-3, which indicates the existence of the B-O stretching vibration (Zhang et al., 2020). The infrared peak centered at 1112 cm⁻¹ belongs to the vibration of SO₄ ²⁻ residue in NCBO/RGO-3, while the infrared peaks located at 671 and 561 cm⁻¹ are derived from the M-B (M = Co, Ni) bonds between the metal and boron atoms (Chen et al., 2017b).

The elemental compositions and binding energy of NCBO/RGO-3 were analyzed by XPS spectrum, as shown in Figures 4C–F. The spectrums of Ni 2p, Co 2p, and C 1s were computer fitted by the Gaussian fitting method. For the Ni 2p spectrum (Figure 4C), there are two satellite peaks at the higher binding energy of Ni 2p_3/2 (856.7 eV) and Ni 2p_1/2 main peaks (873.6 eV), respectively, indicating that the superficial nickel is mainly presented in the oxidation state (Ni²⁺ and Ni³⁺) (Zhang et al., 2019b; Wang et al., 2022). The above Ni²⁺ and Ni³⁺ are derived from nickel borate (labeled as Ni-B_i) (Jiang et al., 2017; Chen et al., 2019b). Similarly, there are also two satellite peaks located approximately 5.5 eV above the major Co 2p_3/2 (781.6 eV) and Co 2p_1/2 main peaks (797.1 eV), respectively (Figure 4D), which are assigned to oxidized Co²⁺ and Co³⁺ species from cobalt borate (labeled as Co-B_i) (Chen et al., 2015; Hu et al., 2019). The distinct peak at 192.4 eV can be attributed to the oxidized boron from the BO₃ ^3- (Figure 2E), indicating that the oxidized boron species are in predominance on the shell of NCBO/RGO-3 (Chen et al., 2017a). Since oxygen is more electro-negative than boron, the superficial metal borides are easily oxidized upon exposure to air, and this phenomenon is also reported in many studies (Masa et al., 2016; Masa et al., 2017). It should be noted that XPS characterization is only sensitive to surface powders (a few nanometers), while it is difficult to analyze the structure of deeper powders. The previous HR-TEM image, XRD pattern, and FT-IR spectrum have demonstrated the existence of Ni-Co borides in NCBO/RGO-3. Therefore, NCBO/RGO-3 can be considered as NCB@NCB_i/RGO composites with Ni-Co boride as the core layer and Ni-Co borate (NCB_i) as the shell layer. These metal borides and borates can provide sufficient active sites for redox reactions, thus effectively improving the electrochemical properties of materials. In C 1s spectrum (Figure 4F), the characteristic peaks at 284.8, 286.4, and 288.9 eV correspond to C-C, C-O, and C=O bonds, respectively. The weak C-O and C=O bonds coupling with strong C-C bond indicate that most oxygen-containing functional groups on GO are removed after calcination, and hence the electrical conductivity of RGO can be greatly enhanced (Zhao et al., 2021a). Therefore, NCBO/RGO-3 is expected to achieve excellent electrochemical performance.

Figure 5A shows the cyclic voltammetry (CV) curves of the three TMBO/RGO samples at the scan rate of 50 mV s⁻¹. A pair of strong and well-defined redox peaks is visible for all three TMBO/RGO, which are mainly attributed to the M-B-related redox reactions. Among them, the CV curve of NCBO/RGO-3 with the largest surrounded area indicates its highest specific capacitance. When the scan rate is increased from 2 mV s⁻¹ to 50 mV s⁻¹, the shape of CV curves for the oxidation and reduction peaks of NCBO/RGO-3 is basically unchanged except for the increase of the respective peaks currents (Figure 5B), signifying that NCBO/RGO-3 electrode has excellent rate capability and electrochemical reversibility (Guo et al., 2020).

The galvanostatic charge-discharge (GCD) curves of the three TMBO/RGO samples at the current density of 10 A g⁻¹ are shown in Figure 5C. It can be seen that the charging and discharging curves of NCBO/RGO-3 are nearly symmetrical and have the longest discharge time among the three TMBO/RGO samples, indicating a good electrochemical reversibility and the highest specific capacitance of NCBO/RGO-3. The GCD curve of NCBO/RGO-3 has an obvious plateau (Figure 5D), which again proves its pseudo-capacitance characteristics (Huo et al., 2019). Figure 5E shows the specific capacitances of the three TMBO/RGO samples at different current densities. At the current density of 1 A g⁻¹, the specific capacitances of NCBO/RGO-3, NBO/RGO, and CBO/RGO are 2383, 1647, and 517.5 F g⁻¹, respectively. When the current density increases to 20 A g⁻¹, NCBO/RGO-3 still maintains 69.3% of specific capacitance and is much better than NBO/RGO and CBO/RGO in terms of specific capacitance and rate capability. In addition, the supercapacitive performances of the as-prepared NCBO/RGO composites are greatly influenced by their Ni/Co molar ratios. It is found that NCBO/RGO-0.5, whose Ni/Co molar ratio is 0.5, displays a specific capacitance of 1415.5 F g⁻¹ that much higher than CBO/RGO (Supplementary Figures S4A–C). With an increase of Ni/Co ratio, the specific capacitances of NCBO/RGO composites gradually increase first and then decrease, and the highest specific capacitance is obtained with NCBO/RGO-3. Additionally, the rate capability of NCBO/RGO-3 is much better than those of NCBO/RGO composites with other Ni/Co ratios when the current density is increased to 20 A g⁻¹ (Supplementary Figure S4C). Therefore, the optimal Ni/Co molar ratio is determined to be 3.

The electrochemical impedance spectroscopies (EIS) of the three TMBO/RGO samples in the frequency range of 0.001 Hz–100 kHz are shown in Figure 5F. The intersection of the semicircular ring on the Z′ axis in the high-frequency region represents the solution resistance R _s of the test system, while the diameter of semicircular ring determines the electron transfer resistance R _ct, and the angle of the slope in the low-frequency region is related to the Warburg impedance. In terms of R _s, NCBO/RGO-3, NBO/RGO, and CBO/RGO are close because all three electrodes are evaluated in the same electrolyte. The electron transfer resistances (R _ct) of NCBO/RGO composites are much smaller than that of NBO/RGO (R _ct = 1.49 Ω), which suggests that the doping of Co elements can significantly promote charge transfer (Figure 5F and Supplementary Figure S4D). In addition, NCBO/RGO-3 has a nearly vertical Warburg angle as compared to CBO/RGO, indicating faster electrolyte ion diffusion within the NCBO/RGO-3 electrode (Zhang et al., 2020). It also has been suggested that the M₃(BO₃)₂ (M = Ni, Co) shell layer promotes the absorption of electrolyte OH⁻ on the electrode surface (Chen et al., 2019b), and the abundant mesopores inside the NCBO/RGO-3 can facilitate the transport and infiltration of electrolyte ions within the whole electrode material. Therefore, NCBO/RGO-3 exhibits excellent specific capacitance and rate capability.

Cycling stability is an important index to evaluate the performance of supercapacitors. The cycling stability of the three TMBO/RGO samples at 5 A g⁻¹ is shown in Figure 5G. For NBO/RGO, significant capacitance degradation is observed with only 56.2% of the initial capacitance remaining after 2500 charge/discharge cycles. The reason is that the loading particles of NBO/RGO are relatively large and poorly conductive, and the electrode material is easily pulverized and then delaminated from the current collector during the fast charge/discharge processes. CBO/RGO possesses low capacitance but good stability with a gradual increment in specific capacitance during 5000 charge/discharge cycles, mainly because the electrolyte ions slowly infiltrates into the internal pores to fully activate the electrode during cycling (Zhang et al., 2019b). In contrast, NCBO/RGO-3 still maintains 97.1% of the initial capacitance after 5000 charge/discharge cycles, which displays a gratifying capacitance, an excellent rate capability and a superior cycling stability that outperform most TMBOs in literatures (Supplementary Table S1). The excellent supercapacitive properties of NCBO/RGO-3 synthesized by TS-IJMR are mainly due to the following reasons: 1) NBO, CBO, and RGO in NCBO/RGO-3 produce a good synergistic effect, and the doping of Co enhances the electrical conductivity of the material; 2) The 2D conductive network of RGO facilitates the electron transfer throughout the electrode; 3) The RGO sheet not only suppresses the material deformation and structural collapse of NCBO-3 during the fast charge/discharge processes but also enables the uniformly dispersed NCBO-3 particles to achieve a high electroactive surface area; 4) The M₃(BO₃)₂ shell layer of NBCO/RGO-3 contributes to the adsorption of electrolyte ions, and abundant mesopores can facilitate the transport and infiltration of electrolyte within the whole electrode.

It can be seen from Supplementary Figure S5 that the supercapacitive performance of NCBO/RGO-3 synthesized in TS-IJMR is much better than that of NCBO/RGO-S synthesized in STR. The supercapacitive performance of electrode material is mainly determined by its morphology, specific surface area, pore size distribution, etc. (Simon and Gogotsi, 2008), which can be further influenced by the different mixing methods of the two reactors. As mentioned above, during the precipitation of nanomaterials in STR, the reactant solutions are added dropwise into the reactor at a certain rate, and then the concentration and pH in STR are homogenized by constant agitation. The feeding rate should not be too fast due to the poor micromixing efficiency of STR (Bhattacharya and Kresta, 2004), hence it usually takes dozens of minutes or even hours to complete the whole precipitation process. However, even if the feeding rate is strictly controlled to achieve a nearly uniform supersaturation in STR by vigorous agitation, the local concentration, pH, and supersaturation are neither spatially nor temporally constant when a drop of reactant solution is added into the STR, in which the previously generated precipitates and dissolved ions have already coexisted. In addition, the newly added metal ions will adsorb on the surface of the previously generated precipitates to form bigger aggregates (Cheng et al., 2009; Alexandrov, 2014). As a consequence, the NCBO/RGO-S synthesized in STR is severely agglomerated (Figure 2F), with a BET specific surface area of only 36.73 m² g⁻¹ (Supplementary Figure S6). This makes substantial particle surfaces cannot be infiltrated by the electrolyte, hence reducing the supercapcitive performance of NCBO/RGO-S. In contrast, TS-IJMR has good micromixing efficiency, and Ni²⁺/Co²⁺ solution can quickly achieve homogeneous mixing with GO sheets at the first micro T-junction. Then, the mixture re-impacts with KBH₄ solution at high velocities to obtain uniformly loaded NCBO particles on GO flakes (Guo et al., 2017). In addition, TS-IJMR provides a more homogeneous and stable environment for both particle nucleation and crystal growth on account of its better process control and enhanced micromixing efficiency. Therefore, the as-prepared NCBO/RGO-3 possesses a large BET specific surface area (62.37 m²/g), quantities of suitable mesopores (2–5 nm) and fast ion transport to participate in redox reactions. As a consequence, the supercapacitive performance of NCBO/RGO-3 is much superior to that of NCBO/RGO-S synthesized in STR.

Although NCBO/RGO-3 displays a high specific capacitance, the operating voltage of 0.4 V still hampers its practical applications for energy storage. In order to broaden the operating voltage of the NCBO/RGO-3, the button-type asymmetric supercapacitor NCBO/RGO//AC is assembled with the NCBO/RGO-3 electrode and AC electrode as the positive electrode and negative electrode, respectively (Figure 6A). The CV curves of NCBO/RGO//AC device at different voltage windows are shown in Figure 6B. When the voltage window is increased to 1.8 V, an obvious hump is observed within the potential of 1.6–1.8 V, suggesting that OER takes place in the positive electrode. Therefore, the optimal voltage window for NCBO/RGO//AC device is determined to be 0–1.6 V. The CV shapes does not change significantly as the scan rates increases (Figure 6C), which signifies the fast charge/discharge responses of NCBO/RGO//AC device. In addition, the nearly symmetrical charge-discharge curves of NCBO/RGO//AC further validates its good charge/discharge reversibility and high Coulombic efficiency (Figure 6D). The specific capacitances of NCBO/RGO//AC are 150.8, 144.8, 129.2, 96.1, and 74.3 F g⁻¹ at the current densities of 1, 2, 4, 10, and 20 A g⁻¹, respectively, coupling with 91.5% capacitance retention after 5000 charge-discharge cycles at 2 A g⁻¹ (Figure 6E). Therefore, the assembled NCBO/RGO//AC device displays a good rate capability and cycling stability.

Energy density and power density are two important indicators of supercapacitor performance. Figure 6F demonstrates that the NCBO/RGO//AC device provides a maximum energy density of 53.3 W h kg⁻¹ at a power density of 800 W kg⁻¹. Even at a high power density of 16,000 W kg⁻¹, it can still achieve an energy density of 26.4 W h kg⁻¹, which is higher than that previously reported TMBO-based asymmetric supercapacitors, such as CoMoO₄/Co-B//AC (Hou et al., 2020), CoB-AC//AC (Hou et al., 2019), NiB _x O _y //AC (Qin et al., 2018), Ni₂B/RGO//AC (Cao et al., 2017), Ni _x B/G//AC (Chen et al., 2019b), Ni-Co-W-B-O/RGO//RGO (Xiang et al., 2017). Finally, a yellow LED was lighted for several minutes with two tandem NCBO/RGO//AC devices. Therefore, the NCBO/RGO-3 electrode materials we synthesized have promising prospects for application in the field of energy storage.

In order to explore the potential applications of NCBO/RGO-3 in electrocatalysis, the OER performance of NCBO/RGO-3 was evaluated. The LSV curve of NCBO/RGO-3 is shown in Figure 7A. At a current density of 10 mA cm⁻², NCBO/RGO-3 can obtain an overpotential as low as 309 mV, which is lower compared to the RuO₂ reference catalyst (338 mV), indicating that NCBO/RGO-3 has better catalytic performance than RuO₂.

Stability is also a key parameter to evaluate the performance of OER electrocatalysts. As shown in Figure 7B, the NCBO/RGO-3 catalyst only slightly decays after 10 h of stable cycling at a constant potential of 1.54 V, with the current remaining above 80%. In contrast, commercial RuO₂ catalysts show a sharp current decay at 10 h. Therefore, NCBO/RGO-3 catalyst has good stability in OER. The NCBO-3 grown on the RGO sheet facilitates the exposure of more active sites on the electrocatalyst surface, thus accelerating the OER reaction at the solid-liquid interface. Moreover, the synergistic effect of the components in NCBO/RGO-3 also plays a facilitating role. Therefore, NCBO/RGO-3 represents an OER electrocatalyst with potential applications.