As illustrated in Figure 1 and Figure S1, 3D GB-Pt scaffolds were synthesized through reduction of fast cryogenic hydrolysates of platinum salts in aqueous solution. Specifically, (NH₄)₂PtCl₆ (5 mg) was dissolved in 10 ml deionized water, forming a pale-yellow solution. After standing for 10 h, the hydrolysis of (NH₄)₂PtCl₆ reaches a dynamic equilibrium as follows [30, 31]:

where the hydrolysis equilibrium could be adjusted via tuning the pH values and the concentration of Cl⁻ during the hydrolysis process (Figure S2). After fast cryogenic treatment (the treatment temperature is -196°C), a yellow foam with equilibrium hydrated [PtCl_6−n(OH)_n]²⁻, NH₄⁺, Cl⁻, and nonhydrolytic [PtCl₆]²⁻ was formed. Then, the foam was reduced at 200°C under a mixed gas of H₂ and Ar with a v/v ratio of 1 : 9 for 2 h, affording 3D GB-Pt scaffolds. In contrast, Pt rods were generated through reduction of the slow cryogenic hydrolysates, since hydrolytic equilibrium of (NH₄)₂PtCl₆ went reversely back during the slow cryogenic process.

To identify the dynamic equilibrium hydrolysates of (NH₄)₂PtCl₆ during our fast cryogenic treatment, UV-vis absorption measurement was conducted owing to its high sensitivity to ligand-to-metal (Cl⁻➜Pt) charge transfer of [PtCl₆]²⁻ and [PtCl_6−n(OH)_n]²⁻ complexes [32, 33]. In principle, by adjusting the pH values of (NH₄)₂PtCl₆ solution from 0 to 12, the different dynamic equilibriums could be achieved, as demonstrated by their UV-vis absorption spectra with different absorbance intensities at ~262 nm (Figure 2(a)), which were originated from the charge transfer involving orbitals with Cl ligand π-character [30, 33]. Notably, while adjusting the pH value to 14, all the (NH₄)₂PtCl₆ was conversed to [Pt(OH)₆]²⁻, Cl⁻, and NH₄⁺, where the hydrolytic equilibrium was completely broken [34]. Through fast cryogenic treatment, the hydrolysates at the different dynamic equilibrium states could be well maintained. This could be clearly demonstrated via their UV-vis absorption spectra (Figure 2(b)). On the contrary, through slow cryogenic treatment (the treatment temperature is -5°C), all the adsorption bands of the hydrolysates at the different dynamic equilibrium states showed the similar absorption intensities (Figure 2(c)). It is suggested that the different levels of hydrolysate [PtCl_6−n(OH)_n]²⁻ were recrystallized to (NH₄)₂PtCl₆ during our slow cryogenic process.

To further evaluate the crystalline phases of fast cryogenic intermediates, X-ray diffraction (XRD) measurement was conducted and is shown in Figure 2(d). Interestingly, in the case of fast cryogenic intermediates, except for the characterization peaks of (NH₄)₂PtCl₆, there are four additional XRD peaks detected at 22.9°, 32.6°, 46.9°, and 58.3° (marked with blue diamonds in Figure 2(d)), indexed to (100), (110), (200), and (211) planes of NH₄Cl (JCPDS no. 72-2378), respectively. The typical TEM (Figure S3a-f) and elemental mapping images (shown in Figure S4) reveal that the product NH₄Cl behaves like a matrix to accommodate the homogeneous dispersion of the fast cryogenic Pt-containing intermediates with sizes of 3-5 nm. Hence, through initial hydrogen reduction treatment, numerous individual Pt nanocrystalline structures could be produced, derived from Pt-containing intermediates. Meanwhile, NH₄Cl would be gradually decomposed and leave gaps between the Pt nanocrystalline structures. With further increase of the reduction time, such gaps allow the infusion between adjacent Pt nanocrystalline structures, generating grain boundary-enriched Pt scaffolds as demonstrated in Figure 3 and Figures S5 and S6. The typical TEM images (Figure 3(b)–3(e) and Figure S7) clearly disclose that these scaffolds are constructed from short Pt crystalline wires with average lengths of ~13 nm and diameters of ~4.2 nm. HRTEM images (Figures 3(d) and 3(e) and Figure S8) show clearly the grain boundaries between two short Pt wires and an obvious interlayer spacing of 0.223 nm, corresponding to d(111) of Pt [35, 36]. In contrast, different levels of hydrolysate [PtCl_6−n(OH)_n]²⁻ and Cl⁻ were recrystallized to single-crystalline (NH₄)₂PtCl₆ during the slow cryogenic treatment, without the detection of equilibrium products of the hydrolysis (Figure 2(d) and Figure S9). After hydrogen reduction treatments of slow cryogenic intermediates, only Pt rods were obtained (Figures S10 and S11). This may be ascribed to the low level of NH₄Cl in the slow cryogenic intermediates, which are unable to prevent the fast growth of big Pt crystalline structures. Moreover, pH values of the (NH₄)₂PtCl₆ solution also affect the formation of 3D GB-Pt scaffolds. As shown in Figures S12 and S13, while the pH value is 2 and 7, the resultant GB-Pt scaffolds show interconnected 3D networks. As the pH value is 0, all the nanowires or particles are strongly aggregated, owing to the less amount of the hydrolysis product NH₄Cl, which cannot efficiently prevent the aggregation during reduction process. In contrast, as the pH value arrives to 12 and 14, all the Pt nanocrystalline structures are separated, without formation of good networks. This should be ascribed to the excessive hydrolysis product NH₄Cl after fast cryogenic treatment, which prevents the infusion between the formed Pt nanocrystalline structures during the reduction process. These results manifest that the appropriate amount of NH₄Cl plays a key role to the formation of 3D grain boundary-enriched Pt scaffolds. To further confirm this hypothesis, two chloroplatinates without ammonium (Na₂PtCl₆ and K₂PtCl₆) were selected to substitute the platinum precursor (NH₄)₂PtCl₆ during our fabrication processes. As shown in Figure S14a-c, only Pt nanoparticles were generated after reduction of the fast cryoequilibrium hydrolysates of K₂PtCl₆. If we deliberately added an amount of NH₄Cl into the K₂PtCl₆ system, 3D Pt scaffolds could be formed again (Figure S14d-f and Figure S15). This phenomenon appears again as Na₂PtCl₆ was used as the Pt precursor (Figures S16 and S17). Therefore, the presence of a moderate amount of NH₄Cl is the essence of the production of 3D grain boundary-enriched Pt scaffolds. In this manner, various grain boundary-enriched metal (e.g., Pd) scaffolds could be fabricated via our fast cryogenic treatment with the presence of a NH₄Cl additive, and the detailed characterizations of 3D GB-Pd scaffolds are shown in the Supplementary Materials (Figure S18-S21).

It should be noted that the grain boundary density of the sample can be well controlled through tuning of the hydrolysis equilibrium by adjusting the concentrations of (NH₄)₂PtCl₆ solution before fast cryogenic treatment. As shown in Figure S22, a higher concentration of (NH₄)₂PtCl₆ leads to an increase of the length of Pt wires. Based on the TEM analysis, we quantitatively measured the average Pt grain length of the GB-Pt scaffolds. It is shown that the average grain length of GB-Pt scaffolds-5 is only 13.2 nm (the grain boundary density of 75.9 μm⁻¹), which is the highest grain boundary density among all the GB-Pt scaffolds samples. (For details, see the Supplementary Materials.) In this manner, the grain boundary density of 3D GB-Pd scaffolds was 64.1  μm⁻¹ (Figure S23).

To gain further insight into the crystalline properties of 3D GB-Pt scaffolds, XRD, XPS, and nitrogen adsorption/desorption isotherm were carried out. As revealed in Figure 4(a) and Figures S24 and S25, in the case of 3D GB-Pt scaffolds, there are three prominent XRD peaks at 39.8°, 46.2°, and 67.7°, corresponding to the (111), (200), and (220) planes of face-centered-cubic (fcc) platinum (JCPDS no. 87-0647) [37], respectively, in good accordance with the above HRTEM and SAED analyses. The XPS survey demonstrates that there are only two main species of Pt and O detected in our 3D GB-Pt scaffolds, without other impurities (Figure 4(b) and Figures S26 and S27). N₂ adsorption-desorption data indicates a high specific surface area of 50.6 m² g⁻¹, much higher than that of the Pt rods (21.1 m² g⁻¹) sample (Figure 4(c)).

The electrocatalytic activities of 3D GB-Pt scaffolds for MOR were investigated directly via cyclic voltammograms (CVs) in an electrolyte of 0.5 M H₂SO₄ and 1 M methanol. As displayed in Figure 5(a) and Figure S28, a high electrochemically active surface area (ECSA) value of 74.8 m² g_Pt⁻¹ is obtained in the case of GB-Pt scaffolds, much higher than in Pt rods (26.9 m² g_Pt⁻¹) and Pt/C (35.9 m² g_Pt⁻¹) (Table S1). Accurately, as shown in Figures 5(b) and 5(c), an ultrahigh peak with current density of 203.8 mA cm⁻² is achieved, corresponding to 1027.1 mA mg_Pt⁻¹ for 3D GB-Pt scaffolds in mass activity, much higher than that of the Pt/C sample (345.2 mA mg_Pt⁻¹) (Figure S29). Moreover, the onset potential of 3D GB-Pt scaffolds is only 217 mV, much lower than those of Pt rods (257 mV) and the Pt/C catalyst (356 mV). Associated with the above TEM and HRTEM analyses, such high electrocatalytic activities of 3D GB-Pt scaffolds should be attributed to the large presence of grain boundaries of Pt. This can be further demonstrated by the different 3D GB-Pt scaffolds with tunable intensities of grain boundaries, in which their activities are linearly proportional to the densities of grain boundaries of Pt (Figure 5(d)). The electrocatalytic activity of 3D GB-Pt scaffolds obtained at different pH values was also systematically investigated (Figure S30). Among them, 3D GB-Pt scaffolds-pH(7) exhibit a very high value of 1027.1 mA mg_Pt⁻¹, much higher than other samples (368.9, 586.9, 485.6, and 166.6 mA mg_Pt⁻¹ for 3D GB-Pt scaffolds-pH(0), (2), (12), and (14), respectively).

To further investigate the electrocatalytic stability of 3D GB-Pt scaffolds for MOR, we conducted a stability test in 0.5 M H₂SO₄ and 1 M methanol. Remarkably, even after 5000 cycles, there is only 9.8% loss of the initial activity for 3D GB-Pt scaffolds (Figure 5(e)). This value is much lower than that of the Pt/C catalyst (66.5% loss of the activity), clearly confirming the excellent durability of 3D GB-Pt scaffolds for electrocatalytic MOR. To gain insight into the reason of the excellent durability performance of 3D GB-Pt scaffolds, CO-stripping measurement was conducted [38]. As demonstrated in Figure S31, the peak potential of 3D GB-Pt scaffolds is only 0.54 V (vs. SCE), much lower than that of Pt/C (0.61 V vs. SCE), suggesting that 3D GB-Pt scaffolds have an enhanced antipoisoning (CO) property [39]. This should ascribe to the presence of large Pt grain boundaries that can efficiently oxidize the adsorbed CO_ads [40]. Furthermore, after durability measurement, both the configurations and grain boundaries of 3D GB-Pt scaffolds were well preserved (Figure S32). The electrochemical impedance spectroscopy (EIS) of the 3D GB-Pt scaffolds, Pt rods, and Pt/C electrocatalysts was carried out to investigate the kinetics of methanol oxidation. As shown in Figures S33 and S34, typical methanol oxidation behaviors catalyzed by the Pt-based catalyst are observed at a series of potentials from 0.1 to 1.0 V [11]. Among them, 3D GB-Pt scaffolds have a much smaller semicircle diameter than Pt rods and Pt/C electrocatalysts at 0.4 V (Figure 5(f)), demonstrating the high methanol oxidation rate of 3D GB-Pt scaffolds [41], in accordance with the electrocatalytic MOR results (Figure 5(b)). In addition, 3D GB-Pd scaffolds also exhibit remarkably high activity of 1117.9 mA mg_Pd⁻¹ (Figure S35) and good stability (9.0% loss of the initial activity after 5000 CV cycles) towards MOR (Figure S36).

Ammonium hexachloroplatinate ((NH₄)₂PtCl₆), potassium hexachloroplatinate (K₂PtCl₆), sodium hexachloroplatinate (Na₂PtCl₆), and ammonium hexachloropalladate ((NH₄)₂PdCl₆) were purchased from Alfa Aesar. Ammonium chloride and sodium chloride were purchased from Beijing Innochem Technology Co., Ltd.

For the synthesis of 3D GB-Pt scaffolds, a certain amount of (NH₄)₂PtCl₆ was initially added in DI water (10 ml) to form a pale-yellow solution. The solution was stirred for 60 min and then frozen at -196°C in liquid nitrogen for 30 min; after that, the obtained frozen product was further dried under vacuum conditions at -60°C for 48 h, and a three-dimensional yellow foam was obtained. The foam was further reduced at 200°C for 2 h under 10% H₂/Ar gas, generating 3D GB-Pt scaffolds. 3D GB-Pt scaffolds-X were fabricated, where X represents the amount of (NH₄)₂PtCl₆ in the 10 ml solution before fast cryogenic treatment. 3D GB-Pt scaffolds-pH(Y) were also fabricated, where Y represents the pH value of the (NH₄)₂PtCl₆ solution before fast cryogenic treatment.

The synthetic procedure of Pt rods was similar to that of 3D GB-Pt scaffolds except that the fast cryogenic treatment was replaced by slow cryogenic treatment (the freezing temperature changed form -196°C to -5°C).

The synthetic procedure of 3D GB-Pd scaffolds was similar to that of 3D GB-Pt scaffolds except that the (NH₄)₂PtCl₆ was replaced by (NH₄)₂PdCl₆, and 5 mg of NH₄Cl should added into the (NH₄)₂PdCl₆ solution before fast cryogenic treatment.

Nanostructures and morphologies of all the samples were carried out through FESEM (JEOL-7500) and HRTEM (JEOL, NEM-2100F). Nitrogen sorption isotherms and BET surface area were measured with Quadrasorb at 77 K. The powder X-ray diffraction patterns were conducted by a Rigaku D/max2500PC diffractometer with a Cu Kα radiation over the range 35-80°. XPS data and chemical bonding nature of Pt and O elements were acquired using Thermo Scientific ESCALAB 250Xi X-ray photoelectron spectroscopy (XPS).

Electrocatalytic properties of our samples were investigated by using an AutoLab workstation in a three-electrode setup cell. Typically, 3 mg of sample and 80 μl of 5 wt% Nafion solution was dispersed in 1 ml of 4 : 1 v/v deionized water/ethanol by sonication, forming a black dispersion. Next, 5  μl of the as-prepared dispersion was casted onto a glass carbon electrode 3 mm in diameter as the working electrode. Meanwhile, a saturated calomel electrode (SCE) was applied as the counter electrode and a Pt sheet was applied as the reference electrode. The typical electrochemically active surface area (ECSA) measurement of our samples was tested in nitrogen-saturated 0.5 M H₂SO₄ electrolyte at 10 mV s⁻¹. The MOR electrocatalytic activity was recorded in 0.5 M H₂SO₄ and 1 M methanol electrolyte at 20 mV s⁻¹. For Pd-based catalysts, the MOR electrocatalytic activity was recorded in 1 M KOH and 1 M methanol electrolyte. The chronoamperometry (CA) test was conducted for a period of 1 h at room temperature. The electrochemical impedance spectra (EIS) were recorded at the frequency range from 100000 Hz to 0.01 Hz with an amplitude of 10 mV.