The field experiment was conducted in Gongzhuling City, Northeast China (124°82′26″ E, 43°79′13″ N). Three typically and locally cultivated varieties with varying protein contents were selected ( Table 1 ). The field experiment was sown on April 28, 2023 and harvested on October 2, 2023. Four nitrogen application rates were 30 kg·hm^-2, 60 kg·hm^-2, 180 kg·hm^-2, and 300 kg·hm^-2. Three ears were randomly selected per plot after harvest, from the middle of each 30 well-developed kernels were sampled. Subsequently, kernels were deactivated by drying at 105°C for 30 minutes at first, and then were dried at 80°C until constant weight was achieved. Ultimately, three kernels with uniform shape and size were randomly selected from the total of 90 for the soaking experiment.

This study extracts relevant indicators from the water phases and phenotypic structure of the maize kernels. The classification of water phases in the kernels is based on the study by (Li et al., 2023), where the relaxation time of bound water ranges from approximately 0.1 to 10 ms (T₂₁ ), semi-bound water from approximately 10 to 100 ms (T₂₂ ), and free water from approximately 100 to 1000 ms (T₂₃ ). The signal amplitudes of the relaxation inversion spectra for bound water, semi-bound water, and free water are denoted as A1, A2, and A3, respectively. A represents the total signal amplitude of the maize kernel, calculated as follows:

Previous studies have shown a significant linear relationship between the water content of maize kernels and the signal amplitude. In this study, the kernels were initially dried, and their dry weight (DW) was recorded. Following each water soaking, fresh weight (FW) of each individual kernel was measured. Water content (TWC) and moisture content (MC) were calculated as follows:

Based on this, the content of each water phase is calculated as follows:

where BWP, SBWP, and FWP indicate the proportions of bound, semi-bound, and free water in the kernel relative to the total water, respectively; BWC, SBWC, and FWC represent the absolute content of bound, semi-bound, and free water in the kernel, respectively. The proportions of the floury endosperm, vitreous endosperm, endosperm, and embryo volumes to the total kernel volume (FEP, VEP, ENDP, and EMBP) are calculated by the following equations:

where FEV, VEV, ENDV, and EMBV represent the volumes of the floury endosperm, vitreous endosperm, endosperm and embryo, and KV represents the total volume of the kernel.

Data processing, statistical analysis and data visualizations were performed using Microsoft Office, SPSS, and RStudio software. Multiple comparisons in this study were performed using Duncan’s SSR test, and correlation analysis was carried out using Pearson correlation coefficients. The significance level was evaluated using p-values, with p< 0.05 considered statistically significant. The calculation method for the linear correlation coefficient is as follows:

where x i   and y i represent the i-th values of x and y, respectively, and x ¯ and y ¯ represent the mean values of x and y, respectively.

The analysis of T₂ relaxation spectra at various soaking durations reveals the distribution and transition of water phase within the maize kernel. The LF-NMR T ₂ relaxation time distribution of maize kernels for different cultivars under different nitrogen treatments soaking with different durations are shown in Figure 3 . The amplitude of signal shows typical pattern with several peaks in maize kernels ( Figure 3 ). With soaking duration increased, these peaks evolved into three distinct categories: bound water, semi-bound water, and free water. Bound water exhibited the strongest signal intensity, followed by semi-bound water, with free water showing the weakest signal. As soaking time increases, the peaks shifted progressively to the right of the x-axis, indicating a transition in water phase, especially as the bound and semi-bound water signals intensify. Both the maximum amplitude of signal and the corresponding relaxation time increased with increasing soaking time across cultivars and nitrogen treatments. The maximum amplitude of signal increased with increasing nitrogen rate. Under low nitrogen rate, the maximum signal was 1328.40 a.u. at 2.97 ms on average for DK517, and 1221.89 a.u. at 3.30 ms for LY99, and 1453.01 a.u. at 2.97 ms for ZD958. Under high nitrogen rate, the maximum signal was 1407.56 a.u. at 2.77 ms on average for DK517, and 1557.60 a.u. at 3.18 ms for LY99, and 1890.64 a.u. at 2.77 ms for ZD958.

The soaking time, cultivar, nitrogen rate significantly affected BWP, SBWP, and FWP ( Table 2 ). Nitrogen rate significantly interacted with cultivars, indicating that the nitrogen effects on kernel water phase were influenced by cultivars. BWP generally increased initially, reaching a maximum in stages T1, T2, or T3, before gradually decreasing ( Figure 4 ). At the T5 stage, the BWP of DK517 under different nitrogen treatments decreased by 9.93%, 7.10%, 9.57%, and 19.20%, respectively, compared to the maximum proportion at the time of submergence, with the greatest decrease at the nitrogen application rate of N300 ( Figures 4A–D ). LY99 showed reductions of 13.58%, 16.33%, 12.10%, and 12.10%, with the highest reduction at N60 ( Figures 4E–H ), while ZD958 decreased by 9.83%, 21.11%, 17.45%, and 14.60%, with the largest reduction at N60 ( Figures 4I–L ). Conversely, SBWP showed an initial decrease, reaching its lowest level during T1 and T2 before gradually increasing. At T5, the SBWP in DK517 increased by 87.53%, 99.12%, 80.85%, and 110.15% from its minimum, with the highest increase at N300 ( Figures 4A–D ). LY99 increased by 69.54%, 105.61%, 98.00%, and 114.01% ( Figures 4E–H ), while ZD958 saw increases of 93.98%, 162.38%, 140.32%, and 121.79%, with the largest gain at N60 ( Figures 4I–L ). The FWP was highest at T1, showing a steady decline over time. By T5, the FWP in DK517 had decreased by 53.88%, 55.75%, 51.50%, and 48.45% across nitrogen treatments, with the most pronounced reduction at N60 ( Figures 4A–D ). LY99’s FWP dropped by 50.86%, 59.42%, 45.85%, and 56.39%, with the greatest decline at N60 ( Figures 4E–H ), while ZD958’s FWP decreased by 54.27%, 52.53%, 49.66%, and 49.04%, with the largest drop at N30 ( Figures 4I–L ). These findings suggest that the predominant water phase in the kernel is bound water. However, with prolonged soaking durations, the proportion of bound and free water decreases, while the proportion of semi-bound water increases.

Maize kernel MC were significantly affected by soaking time (ST), cultivars (C), and their interaction (ST×C) (P<0.01) ( Table 2 ). For DK517 kernels, the average moisture content at successive soaking times was 7.77%, 16.11%, 22.04%, 25.11%, and 29.04%. In LY99 kernels, these values were 8.78%, 17.02%, 23.38%, 26.69%, and 31.03%, respectively, while ZD958 kernels showed mean moisture contents of 6.78%, 13.14%, 18.94%, 23.04%, and 28.43% at each soaking stage ( Supplementary Figure S1 ). BWP for DK517 was significantly decreased by increasing nitrogen rate with a minimal of 61.67%, 58.83%, and 57.22% under N300 in T3, T4, and T5 ( Supplementary Table S1 ). Varietal differences in BWP were primarily significant at T2, with the descending order ZD958 > LY99 > DK517. SBWP for ZD958 was significantly increased by increasing nitrogen rate with a maximal of 21.22%, 31.90%, 35.04%, and 36.31% under N300 in T2, T3, T4, and T5. BWP and SBWP under different nitrogen rate for LY99 and ZD958 was not significantly differed except for T3 for ZD958. LY99 had a significantly higher SBWP than other cultivars under N30 in most soaking durations. FWP was significantly increased by increasing nitrogen rate for LY99 but not affected for DK517 and ZD958.

The soaking time and nitrogen rate significantly affected TWC, BWC, SBWC, and FWC ( Table 2 ). Soaking time significantly interacted with nitrogen rate, indicating that the nitrogen effect on water content in different phases were influenced by soaking time. The water content of maize kernels generally increased with prolonged soaking ( Table 3 ). TWC was significantly increased by increasing nitrogen rate across cultivars ( Table 3 ). BWC was significantly increased by increasing nitrogen rate across soaking durations and cultivars. BWC reaches the maximum value of 90.32 mg, 109.71 mg, and 107.34 mg under N180, N300, and N300 at T5 for DK517, LY99, and ZD958 respectively. SBWC under N300 was significantly higher than that under low nitrogen rate (N30 and N60 at T2, N30 at T3, N30, N60 and N180 at T4, N60 at T5) for DK517, but not differed between nitrogen rate for LY99 and ZD928, indicating the significant interaction between cultivars and nitrogen (P<0.05, Table 3 ). Compared to TWC, FWC showed a similar pattern of response to nitrogen across cultivars, with peaking at N300 with values of 9.92 mg, 8.65 mg, and 9.53 mg for DK517, LY99, and ZD958, respectively.

The KV and EMBP were found to be significantly affected by cultivars, nitrogen rate and their interaction ( Table 2 ). FEV was significantly influenced by cultivars and its interaction with nitrogen rate. VEV was significantly influenced by nitrogen rate. ENDV was significantly affected by both cultivars and nitrogen rate. KV under high nitrogen rate (811.97 mm³, 798.37 mm³, 857.43mm³) was significantly greater than that under low nitrogen rate (677.38 mm³, 557.29 mm³, 636.07 mm³) for DK517, LY99, and ZD958 ( Table 4 ). ENDV and EMBV showed similar patterns with values of 597.18 mm³, 80.20 mm³ for DK517, 501.16 mm³, 56.12 mm³ for LY99, and 565.59 mm³, 70.48 mm³ for ZD958 under low nitrogen rate, and that of 723.46 mm³, 88.51 mm³ for DK517, 711.89 mm³, 86.48 mm³ for LY99, and 752.60 mm³, 104.82 mm³ for ZD958 under high nitrogen rate. FEV was significantly increased by increasing nitrogen rate for DK517 and LY99, but not for ZD958. VEV was significantly increased by increasing nitrogen rate for LY99 and ZD958, but not for DK517. Only VEP was significantly affected by cultivars, while other proportions were not significantly affected by neither the cultivars nor nitrogen rate, suggesting a stable spatial distribution inside maize kernel.

A strong correlation was found between water phase distribution and the internal structure of kernels ( Figure 5 ). A significant positive correlation was observed between A and WC, and between NA and MC in maize kernels ( Supplementary Figure S2 ), demonstrating the efficacy of LF-NMR in measuring water content of individual kernel. BWC, SBWC, and FWC showed significantly positive correlations to VEV, ENDV, EMBV, and KV, suggesting a strong relationship between water content in difference phase and internal structure of maize kernel. BWC and SBWC correlates most strongly with KV (0.85, 0.41), followed by ENDV (0.84, 0.41), VEV (0.77, 0.37), and EMBV (0.67, 0.34). However, FWC correlates most strongly with KV (0.80), followed by ENDV (0.78), EMBV (0.72), and VEV (0.65). MC was found to correlate most strongly with NA2 (0.90), followed by SBWP (0.87), NA (0.86), SBWC (0.81), and A2 (0.75), suggesting that signal of semi-bound water might be a robust indicator to predict moisture content in maize kernel. SBWP showed a linear relationship with MC across soaking time, but the relationship was stronger for higher MC levels (R² = 0.69-0.82 for MC > 10%) than lower MC levels (R² = 0.31 for MC< 10%) ( Figure 6 ). The significantly negative correlations between BWP and SBWP (-0.99), SBWC (-0.85), and MC (-0.83) demonstrate a clear trend of increasing semi-bound water proportion with elevated water content ( Figure 4 ).

Visual comparison between water distribution and internal structure in individual kernel for different cultivars under different nitrogen rate were shown in Figure 7 . The T₂ signal in the embryo area was substantially stronger than other tissues, showing a gradually diminishing pattern from the base towards surrounding tissues across all cultivars and nitrogen treatments. This indicates that water may infiltrate the kernel via embryo and then gradually distribute into floury endosperm. The signal within the embryo region increased substantially with increasing nitrogen application rate across all cultivars. DK517 exhibited a significantly higher signal at the kernel top under high nitrogen, suggesting a different water distribution pattern outside the embryo compared to LY99 and ZD958.

The LF-NMR technique is a method based on the spin properties of the hydrogen nucleus in a magnetic field, which makes it possible to jump from a high energy state to a low energy state in a non-radiative way. By measuring the absorption of the resonant radio frequency by the nuclear spin of the proton, LF-NMR can monitor proton states with different relaxation times and their distribution (Carneiro et al., 2013). Because hydrogen protons are abundant in water molecules, LF-NMR analysis of relaxation time changes allows for a quantitative and qualitative study of water content, distribution, and migration within materials (Kamal et al., 2019). Meanwhile, there was a good correlation between the water content of the samples during the soaking and the amplitude of the T₂ inversion spectrum signal in LF-NMR ( Supplementary Figure S2 ), affirming the reliability of LF-NMR for examining maize kernel water status. Notably, signal peaks of each water phase shifted rightward with longer soaking times, and peak intensities increased ( Figure 3 ), suggesting patterns of water migration similar to those found in other crops. Studies on rice have demonstrated similar rightward shifts in both bound and free water peaks with prolonged soaking (Wang et al., 2024), suggesting that water dynamics during absorption may be consistent across different crops. Additionally, the position of the T₂ relaxation inversion peak in LF-NMR reflects the kinetic properties of kernel moisture, particularly the degree of binding between water molecules and macromolecules (Mou et al., 2016).

The soaking process of maize kernels conforms to Fick’s law of diffusion, where the diffusion rate gradually decreases over time until equilibrium moisture content is reached (Antwi et al., 2022). Studies have shown that temperature significantly affects water absorption rates (Nweke, 2010). However, this study excluded temperature as a variable and focused on the impact of kernel structure on water absorption behavior in different maize cultivars. This study found that the initial water absorption rates of different maize cultivars (e.g., DK517, LY99, and ZD958) were 2.88%.h^-1, 4.12%.h^-1, and 3.18%.h^-1, respectively, during the first 3 hours ( Supplementary Figure S1 ). McDonald et al. (1994) reported that kernels exhibit rapid water uptake within the first 6 hours, and our findings are consistent with this result. Additionally, the high-protein maize cultivar LY99 exhibited higher kernel moisture content compared to other cultivars ( Table 1 ; Supplementary Figure S1 ). This is likely because smaller kernels generally have a larger surface area-to-volume ratio, enabling faster water absorption, while kernels with higher protein or starch content exhibit stronger water-binding capacities (Schopf and Scherf, 2021; Tarahi et al., 2022). This highlights the critical role of water absorption pathways and internal structure in maize kernels.

The mechanism of water uptake by maize kernels during soaking is a complex process involving both physical and chemical changes. It is widely accepted that water initially penetrates the pericarp, then moves into the endosperm, and eventually diffuses into the embryo. Agarry et al.(2014) reported that water uptake primarily occurs through capillary action in the outer pericarp, resulting in a rapid increase in kernel moisture. Our LF-NMR visualization further confirmed this, showing strong signals in the embryo region ( Figure 7 ), which aligns with the findings of Chen et al. (2020). Similarly, Miano et al. (2017) suggested that moisture first reaches the embryo and then spreads to the endosperm during early stages of soaking. Additionally, Meyer et al. (2007) and Penfield (2017) observed similar water absorption pathways in soybean and rapeseed. Thus, LF-NMR proves to be an effective tool for tracking water migration within kernels. Additionally, we observed a positive correlation between kernel volume and water content, likely due to larger seeds having a greater surface area for water absorption. The significant positive correlation between ENDV and WC in this study also suggests that the endosperm, as a major storage site for macromolecules like starch and proteins, plays a crucial role in binding water and enhancing water uptake. This finding is consistent with previous studies (Uriarte-Aceves and Sopade, 2023).

The process of water soaking in maize kernels is mainly driven by osmotic pressure differences, where the osmotic pressure of the external water is lower than that inside the kernel, prompting the movement of water towards the interior of the kernel. In the present study, it was found that kernel moisture increased with duration of soaking until osmotic pressure stabilized ( Supplementary Figure S1 ). The loose structure of the endosperm facilitates the rapid entry of water into the cell (Wen et al., 2017), providing a physical basis for water uptake and distribution. This is supported by the fact that the internal structural properties of the kernel were significantly correlated with water phase distribution ( Figure 5 ). Bound and semi-bound water are easily bound to macromolecules and are usually found in the nutrient-rich regions of the kernels (Nuss and Tanumihardjo, 2010; Xu et al., 2019b; Wang et al., 2019). The correlation between bound water and horny endosperm was higher than that of powdered endosperm, probably because the protein content and spatial distribution were related to horny endosperm (Xu et al., 2019b), reflecting the interaction between water and nutrients. During the soaking process, the proportion of bound water gradually decreased, and the proportion of semi-bound water increased, indicating that water began to enter the cell interior. During this process, the water entering the kernel interacts with macromolecules. Once the bound water has fully interacted with the macromolecules, the additional water entering the kernel exists in a state between free water and bound water, leading to an increase in the proportion of semi-bound water. The small increase in the proportion of free water may be due to the entry of water into the voids and cracks of the kernels through capillary action. The moisture signals of the embryo were stronger than those of the chalky endosperm and horny endosperm, which was further confirmed by the significant correlation between the horny endosperm and the moisture of the kernels ( Figures 5 , 7 ). In conclusion, by exploring the distribution characteristics of kernels structure and water phase, we can reveal the migration characteristics of moisture in the kernels and provide new perspectives for the study of grain quality.

Our results are consistent with the existing literature, particularly with regard to the effect of nitrogen application on water status. The BWP increased with higher nitrogen applications in the study by Krishnan et al. (2014). However, we observed that the ratio of semi-bound to free water was not fully consistent with the previous study with increasing nitrogen application. This may be related to present study of the soaking process of the dry kernels in reverse order, highlighting the differences that can occur under different experimental conditions. For instance, Chen et al. (2020) examined kernel dehydration, water uptake, and rehydration, focusing on dehydration, whereas this study’s focus on soaking highlighted differences in water dynamics. By comparing the water uptake in kernels across different maize cultivars, this study aims to elucidate the response characteristics of water phase in kernels to internal structure and nutrient elements. Overall, these detailed observations on maize kernel water phase dynamics provide insights into water content variability under nitrogen treatments, offering a foundation for similar investigations in other crops.

Nitrogen treatments significantly impact the content of various water phases in maize kernels ( Table 2 ). Mou et al. (2016) found that bound water content indirectly reflects dry matter content in wheat, as water phase distribution is closely tied to soluble sugar and starch content. Similar connections between kernel water phases and nutritional components have been observed in maize (Duan et al., 2024). In this study, absolute water content in each phase is influenced by kernel moisture and macromolecule-water binding strength. Increased nitrogen application leads to higher nitrogen accumulation, lipid, and protein contents in kernels (Hammad et al., 2023; Feng et al., 2024). These findings suggest that nitrogen-induced differences in water phase content stem from changes in kernel internal composition. The higher absolute content of bound water under high nitrogen conditions ( Table 3 ) aligns with findings by Krishnan et al. (2014). This study validates the reliability of using water phase distribution analysis to explore internal kernel components and demonstrates the potential of LF-NMR in evaluating maize kernel quality.

Our study focuses on water phase transitions and their interactions with kernel structures during soaking experiments. While the soaking process differs from natural dehydration, the findings provide valuable insights into water dynamics within kernels. For example, the transformation of bound water to semi-bound water observed in this study reflects the interaction between water molecules and macromolecules, which may also occur during natural dehydration as kernels lose moisture through evaporation and internal transport systems (Zhang et al., 2022). To our knowledge, this study is the first attempt to employ LF-NMR and μ-CT techniques to investigate moisture phase distribution in maize kernels and confirm the feasibility of exploring the interactions between water dynamics and internal structure in maize kernels. However, several limitations exist and need to be addressed in future research.

First, the internal structural dependence of water phase in maize kernel was drawn from soaking kernels for various duration, but whether it is still held for natural dehydration process of maize kernel remain unknown. Future attention should be paid on the relationship between water phase distribution and internal morphological structures during kernel filling. Second, the distribution of different water phases was inferred from the amplitude of correlation coefficient and the spatial visualization of T₂ signal of total water within a kernel ( Figure 7 ). The distribution of water phases and metabolic substances needs to be investigated in detail by 3D-MRI and hyperspectral microscope imaging techniques (Zhang et al., 2024). At last, kernel moisture changes are often quantified through empirical modeling with limited underlying mechanisms (Chazarreta et al., 2023). Computational biology models, such as metabolic transport and water diffusion model, should be integrated to gain a comprehensive understanding of kernel development.