The experiment began in 2015 at C-Block, ZARS, V.C. Farm, Mandya (located at latitude 19° N, longitude 76° E, and 695m altitudes). The soil is classified as Alfisol (Hyperthermic mixed typic Rhodustalf). The textural class is sandy loam with 76.54% sand, 6.56% silt, and 16.90% clay. The pH of the experimental soil (2015) was 7.1 (1:2.5 soil-water suspension) and the electrical conductivity (EC) was 0.22 dS m^-1. At 0-15 cm soil depth, OC content was 11.70 g kg^-1, available nitrogen (N) was 340.48 kg ha^-1, available phosphorus (P) was 203.32 kg ha^-1, available potassium (K) was 348.09 kg ha^-1, 0.5M acetic acid extractable Si (AASi) concentration was 73.82 mg kg^-1, 0.01M extractable CaCl₂ Si (CCSi) concentration was 41.98 mg kg^-1 and 1% Na₂CO₃ extractable amorphous Si (ASi) was 4.91 g kg^-1.

In 2015, the experiment began with the rice-rice cropping system to monitor the inputs, outputs, and the internal silicon transfer in the paddy field to estimate the solute mass balance, status of silicon and bioavailable silicon (Riotte et al., 2018). In 2015, the experiment was started with four replicates of each treatment: (1) barren soil, (2) barren soil with the application of 30 g m^-2 of diatomaceous earth (DE) from Agri Power Australia Ltd., (3) planted with rice, and (4) rice planted with the application of 30 g m^-2 of DE before planting. All plots received the recommended dose of fertilizers (RDF) (100:50:50, N: P₂O₅: K₂O kg ha^-1) as per the package of practices. The entire P (diammonium phosphate, DAP, 18% N and 46% P₂O₅) and K (muriate of potash, MOP, 60% K₂O) dose was applied as basal treatments, while N (urea) was applied in two split doses 30^th and 60^th day after transplanting. The experiment was repeated in 2016 with rice hull biochar at the rate of 2 t ha^-1 as the Si source. In 2015 and 2016, KMP-101, locally called Thanu variety, was used for transplanting. However, from July (Rainy season) 2017, the objective of the experiment was changed to assess the effect of different Si sources on rice. The experimental plots were converted to the aerobic rice plot, and the variety grown was changed to BI-33 (Anagha). Three different Si sources were used i.e., diatomaceous earth (DE), concentrated soluble silicic acid (SA), and rice husk biochar (RHB). The composition of DE and RHB is shown in Table 1 , and concentrated soluble silicic acid (SA), obtained from ReXil Agro BV, Chennai, India, contains 2% Si as soluble H₄SiO₄ (ShwethaKumari et al., 2021). As a result, the treatments were modified: (1) T1: Control (RDF alone); (2) T2: RDF + 300 kg DE ha^-1; (3) T3: RDF + 4 mL SA L^-1; and (4) T4: RDF + 4 t RHB ha^-1. The three Si sources RHB, DE and SA were applied at the rate of 4 t ha^-1, 300 kg ha^-1 and 4 mL L^-1, respectively. The entire dose of DE and RHB was applied as a basal before sowing, while the SA was foliar sprayed every 15 days up to reproductive stage of the rice. To avoid the drift to the adjacent plot while spraying, foliar spray was conducted in the early morning with the help of a knapsack sprayer (AGRIMATE, AM 505E) of 20 L capacity. The application rates of DE and RHB fall within the range of field application rates of Si fertilizers (Sandhya and Prakash, 2018) and biochar (Liu et al., 2013), respectively.

Plant samples were collected from the field after the crop was harvested in June 2019. The samples were washed with deionized water, dried in an oven at 70°C, and grounded with a ball mill (MM 400, Retsch, GmBH, Germany) for further analysis.

After crop harvest, surface soil samples were collected from each plot with a screw auger. Soil samples were air-dried, powdered, and sieved with a 2 mm sieve. The soil pH was measured using a glass electrode with a soil-to-water ratio of 1:2.5 (m: v) (Systronics pH system 362, India). Electrical conductivity (EC in soil water suspension of 1:2.5) was measured with Systronics Conductivity Meter 306, India. The wet oxidation method (Walkley and Black, 1934) was used to analyze the soil OC. The available nitrogen (N) was determined by the alkaline potassium permanganate method (Subbiah and Asija, 1956). Estimation of available phosphorus (P₂O₅) by Olsen’s method (Olsen, 1954). Available K was extracted by using 1 N ammonium acetate and was analysed using flame photometry (Jackson, 1973).

In this study, we extracted three readily soluble Si pools using different extractants: 1) 0.01 M CaCl₂ (Haysom and Chapman, 1975), 2) 0.5 M acetic acid (Narayanaswamy and Prakash, 2009) and 3) amorphous Si was extracted using 1% Na₂CO₃ (Majumdar and Prakash, 2020b). Throughout the paper, Si extracted by 0.01 M CaCl₂, 0.5 M CH₃COOH, and 1% Na₂CO₃ is referred to as CCSi, AASi and ASi, respectively. The detailed extraction procedures are given below.

The data presented in the study is the average of the four replicates. All datasets were analyzed statistically using one-way ANOVA with SPSS and Star software packages at a significance level of P<0.05 via Tukey test. Redundancy analysis (RDA) was carried out to determine the interrelationship between plant phytolith and soil parameters. RDA is an intriguing extension of PCA in which the response variables are explicitly modelled as a function of explanatory variables. In this study, the phytolith and PhytOC parameters were considered as response variables, whereas soil parameters (pH, EC, OC, available N, P, K, CCSi, AASi and ASi) were explanatory variables. To determine the structure and direction of the interdependences among the response variables a PCA was performed using R x 64 4.0.1 software using the ggplot2 package. The correlation matrix was used to extract principal components. Kaiser-Meyer-Olkin (KMO) and Bartlett’s tests were used to determine the suitability of the data for PCA (Maurya et al., 2021). KMO test results greater than 0.50 and Bartlett’s significance level p<0.05, indicated a significant relationship between different variables. The RDA’s graphical output is a triplot, which is made up of two biplots stacked on top of each other. The graph’s three components are 1) explanatory variables, 2) response variables, and 3) treatment. Monte Carlo permutation test with PERMANOVA with 999 permutations is used to determine the order of importance of explanatory variables (adonis function). The F statistic and p value show that the null hypothesis can be rejected. The Monte-Carlo test indicates it is significant.

The content of phytolith significantly increased compared to the control (RDF alone) with application of different Si sources ( Figure 2 ). The different Si sources increased the phytolith content of each rice organ and follow the trend: husk > leaf > straw > grain. The average phytolith content in all the treatments varied significantly from 1.83 to 13.80% among different rice organs. The content of phytolith in leaf (8.45 ± 0.59%) and straw (6.91 ± 0.55%) was significantly (p< 0.05) higher than that of the control. Among the treatments, application of 4 t RHB ha^-1 had the highest phytolith content (8.45 ± 0.59%) in leaf, followed by 4mL SA L^-1 (8.02 ± 0.64%) and 300 kg DE ha^-1 (7.45 ± 0.34%). Phytolith content in straw was recorded to be higher with the treatment receiving 4 t RHB ha^-1 (6.91 ± 0.55%) and the remaining treatments were on par with each other. In the husk, a significantly higher phytolith content of 13.80 ± 0.67% was recorded with the application of 4 t RHB ha^-1, whereas DE application at 300 kg ha^-1 (12.55 ± 0.30%) and SA application at 4 mL L^-1 (12.28 ± 0.48%) produced comparable results. Phytolith content in grain follows the same trend with treatment receiving 4 t RHB ha^-1 recording significantly higher phytolith (3.41 ± 0.16%) followed by 300 kg DE ha^-1 (2.73 ± 0.39%) and 4 mL SA L^-1 (2.61 ± 0.37%).

There was a significant increase in the C content in phytolith (%) and PhytOC (%) with the application of different Si sources ( Figures 3 , 4 ). However, it does not follow any trend within different rice organs and different sources of Si. The C content in the phytolith and PhytOC content in different rice organs ranged from 0.19 to 3.04% and 0.004 to 0.46%, respectively. Different Si sources significantly increased the C content in Phytolith (%) in leaf over control (RDF) but there was no significant difference among the treatments. Treatment receiving 300 kg DE ha^-1 recorded higher C content in straw phytolith (1.48 ± 0.45%) and husk phytolith (0.84 ± 0.05%) ( Figure 3 ). The C content in grain phytolith was found to be higher in the treatment receiving 4 t RHB ha^-1 (0.33 ± 0.01%) and lower in the control (0.19 ± 0.025%) ( Figure 4 ). Application of 4 t RHB ha^-1 had significantly higher PhytOC content in leaf (0.46 ± 0.02%) and grain (0.011 ± 0.001%), while the control (RDF alone) recorded the lowest ( Figure 4 ). Whereas PhytOC content in straw and husk did not differ significantly with the application of different sources of Si ( Figure 4 ).

There was no significant difference in pH and EC among the treatments ( Table 2 ). The average OC content in post-harvest soil ranged from 5.27 ± 0.66 to 6.79 ± 0.63 g kg^-1 (4 t RHB ha^-1 = 4 mL SA L^-1 > 300 kg DE ha^-1> Control (RDF)) recorded significantly (p<0.05) higher organic carbon (OC) ( Table 2 ). The available N content in soil varied from 145.60 ± 7.92 to 198.80 ± 5.60 kg ha^-1 and showed the trend: 300 kg DE ha^-1 > 4 mL SA L^-1 > 4 t RHB ha^-1 > Control (RDF) ( Table 2 ). Available P₂O₅ content varied significantly between 190.24 ± 8.79 to 276.10 ± 14.79 kg ha^-1 and followed the trend: 300 kg DE ha^-1 = 4 t RHB ha^-1 > 4 mL SA L^-1 > Control (RDF) ( Table 2 ). The average available K₂O varied from 381.33 ± 10.48 to 414.03 ± 10.27 kg ha^-1 (4 t RHB ha^-1 = 4 mL SA L^-1> 300 kg DE ha^-1> Control (RDF)) ( Table 2 ).

With the application of different silicon sources, the content of 0.5 M acetic acid extractable Si (AASi) and 0.01 M calcium chloride extractable Si (CCSi) in the soil increased significantly (p<0.05). Treatment receiving 4 t RHB ha^-1 resulted in significantly higher AASi (146.56 ± 6.52 mg kg^-1), CCSi (57.99 ± 17.01 mg kg^-1) and ASi content (5.21 ± 1.00 g kg^-1) values, while RDF resulted in the lowest ( Figures 5 , 6 ). Application of 300 kg DE ha^-1 and 4 mL SA L^-1 produced comparable results.

In this study, we conducted a redundancy analysis to explore the relationships between phytolith and PhytOC content in plants and readily available soil Si pools and soil properties, and how those relationships are affected by different Si sources. Plant parameters (phytolith, C content in phytolith and PhytOC) were considered response variables. Whereas, soil parameters (pH, EC, OC, available N, P, K, CCSi, AASi, and ASi) were explanatory variables. RDA is the multivariate (meaning multi-response) technique analogue of regression. The technique employs a combination of linear regression and principal component analysis (PCA). A PCA analysis was performed on the response variables to generate a set of principal component vectors for the RDA analysis. RDA uses a mix of linear regression and principal component analysis (PCA). To conduct the RDA analysis, a PCA analysis was performed on the response variables to get a set of principal component vectors. Response variables were quantified using principal component analysis and classified as strong, moderate, or weak based on loading values greater than 0.75. Two factors (Principal Components, PC) were extracted that explained 87.3% of the total observed variation in plant phytolith and PhytOC content Eigenvalues greater than 1.0. ( Table 3 ). PC1 explains 62.4% of the variance and was correlated (Factor loading > 0.75) with the phytolith content in the leaf, straw, husk and grain ( Table 3 , Figure 7 ). PC2 explains the 24.9% variance with the strong loading of the C content and PhytOC content in the leaf, straw, husk, and grain. RDA1 and RDA2 total explained proportions were 58.03 and 10.99%, respectively ( Figure 8 ). The model’s significance was determined using Monte Carlo permutation tests (999 permutations). The RDA revealed that the concentrations of OC (F = 5.26, p = 0.06), available N (F = 12.36, p = 0.004***), available K (F = 6.91, p = 0.02*), CCSi (F = 8.16, p = 0.01*), and ASi (F = 44.96, p = 0.001***) were the most important factors influencing the phytolith and PhytOC parameters, as evidenced by significant correlations ( Table 4 ).

In our study, the phytolith content, C content in phytolith and PhytOC content in different rice organs increased with the application of different sources of Si ( Figures 2 – 4 ). Variation in silica deposition in different rice parts is caused by passive transport i.e., organs with higher transpiration rates accumulate more Si (Kumar et al., 2017). On the other hand, Yamaji et al. (2015) found that despite having a lower transpiration rate, the rice husk accumulated more silica due to the presence of two developed vascular bundles below the panicle. The higher Si flow in the husk gives the hard outer coating of grain which protects grain during development. Our findings are consistent with the findings of Gallo et al. (1974), who studied rice, oat, rye, and wheat and reported that the seed coat accumulates the most silica and the grain accumulates the least. A significant positive correlation among phytolith, C content in phytolith and PhytOC have been reported by various researchers (Anjum and Prakash, 2021; Rehman et al., 2023). PhytOC content in plants depends on the phytolith content and the C encapsulation potential of the phytolith during the formation (Anjum and Prakash, 2021). Among the different sources of Si, biochar performed better because plant-derived phytoliths are concentrated in biochar (Li et al., 2020b). Pyrolysis of Si-rich crop residue enhances solubility of phytoliths and their ability to release PASi (Li et al., 2020b), Si uptake and crop biomass (Li et al., 2019b; Li et al., 2019c). Increased PASi with Si fertilizer and biochar application indirectly increased leaf C content in phytolith, improving PhytOC accumulation of Moso bamboo, particularly in treatment receiving biochar (Huang et al., 2020). Guo et al. (2015) found that basaltic powder (BP) increased the phytolith and PhytOC due to increased rice silicon uptake, which accelerated the chemical weathering of primary silicate minerals in basalt and thus the release of dissolved silicon. Li et al. (2020a) found that applying Si-P fertilizer in Si-deficient soils has no discernible effect on phytoliths and phytOC. Sun et al. (2019) conducted a pot experiment in Si-enriched and Si-deficient soils and reported that slag-based Si fertilizer increased phytolith and PhytOC concentration under both conditions. Thus, the external supply of Si through fertilizer and biochar has an impact on phytolith accumulation and PhytOC concentration.

Application of RHB and Si fertilizers had no significant effect on soil pH and EC over control (RDF) ( Table 3 ). Although there was an increase in the soil pH compared to initial values, EC had no definite trend. The initial pH of the experimental soil in the year 2015 was 7.10 but due to continuous monocropping, the pH of the experimental field has increased to 8.2. Xiang et al. (2009) reported that an increase in soil pH from 6.4 to 7.1 after 12 seasons of aerobic cultivation as the soil reduction-driven moderation or amelioration in soil pH observed in submerged rice soils is not observed in aerobic soils. Similar results with the application of DE (Sandhya and Prakash, 2018), calcium silicate, rice, hull and rice hull ash (Sandhya and Prakash, 2017) and biochar (Wu et al., 2021) have been reported in neutral or alkaline soil. There was a significant increase in the SOC content in the treatment receiving 4 t RHB ha^-1 over control by ~ 29% ( Table 3 ). Improvement in available nitrogen (N), phosphorus (P) and potassium (K) was found with the application of different sources of Si ( Table 3 ). Similar results were reported by Sandhya and Prakash (2017). Increase in available N status in soil caused by N and Si fertilization could be attributed to high adsorption capacity of Si and increased microbial activity, which accelerated the mineralization process during the crop growth period, resulting in a high N accumulation in Si-treated soils. Similarly, Yogendra et al. (2017) found that using CaSiO₃ increased N-use efficiency in aerobic rice. In the current investigation, treatment receiving biochar increased SOC significantly, but available N did not follow the same trend, which could be attributed to a high C: N ratio in crop residue-derived biochar (RHB). Similar results were reported by Han et al. (2020) where biochar amendment led to a significant decrease in dissolved organic nitrogen (DON) concentration in both low-fertility and high-fertility soil. Also, slight decrease in the available N several folds from initial year of experiment 2015 (340.80 kg ha^-1 to summer 2019 (145 to 180 kg ha^-1). Because of higher nitrogen losses in form of nitrous oxide (N₂O), through coupled nitrification-denitrification processes. Improvement in the available phosphorus (P) with the application of Si over control can directly increase the available P by reducing Fe-P retention capacity under higher PASi which converts slightly soluble phosphate into plant-available forms (Schaller et al., 2019). Increase in available K₂O content of the soil with the application of Si could be due to the utilization of native K₂O with increasing levels of silicon which resulted in the building up of higher soil K₂O status. The increased available K with Si application could be ascribed to the alterations in crystal structures of clay lattices which might have been related to a reduction in K fixation thus releasing potassium. The application of biochar significantly increased the available K₂O due to a large amount of K absorbed on the surface of biochar are bioavailable and increased activity of K-dissolving bacteria (Xia et al., 2022). The contents of available P and K were increased during the cultivation process (from the year 2015-2019). This was likely due to fertilizer addition during all cultivation seasons. Together, these results suggest that application of Si fertilizer and RHB improves the soil physicochemical properties.

Addition of phytolith-rich biochars in cultivated soils promotes the release of PASi in the soil-plant system (Li and Delvaux, 2019a), therefore, RHB application resulted in higher CCSi and AASi ( Figure 5 ). The application of the biochar produced at a lower pyrolytic temperature (700°C) phytolith exposure is more, as a result, higher dissolution of phytolith (Xiao et al., 2014; Li and Delvaux, 2019a). According to the Si dissolution kinetics, high Si biochar could be a novel slow-release source of biologically available Si in low Si agricultural soils (Xiao et al., 2014). The increased available Si was anticipated with the application of DE as the added DE might have contributed to the release of soluble Si (Sandhya and Prakash, 2018; Sandhya and Prakash, 2019).

The ASi content in this study was analyzed as per Majumdar and Prakash (2020b), which was optimized for the estimation of ASi content in tropical soils. In this study, the soil samples were analyzed in duplicate for each treatment per replication for 3 to 6 h. A regression model was prepared from 3 to 6 h has significant variables and produces a higher value of the coefficient of determination (R²) ( Table S1 ). ASi content in this study ranged from 2.36 to 5.21 g kg^-1 ( Figure 6 ), which was found to be consistent with the previous studies (Guntzer et al., 2012; Nguyen et al., 2019; Majumdar and Prakash, 2020a). Because of the presence of abundant phytolith, the treatment receiving RHB had an empirically higher ASi content. Li et al. (2019c) and Li et al. (2020b) found that using Si (+) biochar resulted in a higher ASi in comparison with Si (-) biochar and wollastonite. Similarly, Majumdar and Prakash (2020b) reported a higher ASi content in the sugarcane soil profile than the rice soil due to the incorporation of burnt sugarcane trash into the soil during harvesting. ASi content in the control (RDF) was lower due to the continuous removal of rice straw from the field. Because, rice cultivation intensifies desilication as rice straw contains 86% of Si taken up by rice plants (Anjum and Prakash, 2021). Dissolution rate of ASi is several orders of magnitude higher than those of kaolinite or plagioclase at near-neutral pH (Riotte et al., 2017). Si isotopic signatures study conducted by Riotte et al. (2017) confirmed that the ASi plays an important source of Si in rice. Guntzer et al. (2012) reported that the soil ASi pool (0.5 mg g^-1 to 4.3 mg g^-1) in Broadbalk, Rothamsted (England) depleted with the export of wheat straw from 1883 to 1944, particularly in surface soil. Desplanques et al. (2006) estimated that irrigation waters bring more than one-third of the Si needed for rice. They conducted solute mass balance study in the rice field of Camargue (France) where DSi in irrigation water was low and rice straw was removed from the field. They hypothesized that ASi being the only source of dissolved silicon, will be exhausted in 5 years. Since our experiment was a 5-year-old intensive rice experiment plot (10 seasons) converted to the aerobic rice experiment for the past 2 years (4 seasons) the risk of ASi depletion in the soil is very high.

PCA data showed that the plant phytolith and PhytOC parameters under investigation were divided into four major groups ( Figure 7 ). According to the ordination of variables in the PCA diagram, variables in control (RDF) appeared to not correlate with phytolith and PhytOC content, whereas the external Si application appeared to have a positive correlation. Among the treatments, the phytolith and PhytOC parameters of the treatment receiving 300 kg DE ha^-1 and 4 mL SA L^-1 clustered together. In this study, the application of biochar increased the plant phytolith and PhytOC content over other Si fertilizer treatments. Whereas, Huang et al. (2020) found that biochar had no effect on PhytOC content in the Moso bamboo plantation but did influence leaf phytolith formation. Several studies have previously found that biochar improves phytolith concentration (Si uptake) in cotton (Li et al., 2018), wheat (Li et al., 2019c) and rice (Wang et al., 2019). However, the impact was caused by variations in vegetation type, source and rate of biochar and PASi (Li and Delvaux, 2019a; Wang et al., 2019).

Monte Carlo permutation tests revealed significant relationships between the concentrations of plant phytolith and PhytOC as well as SOC, available N, available K₂O, CCSi and ASi indicating close association with each other ( Table 4 ). Figure 8 shows that the RDA ordination of variables revealed that the samples of treatments receiving 4 t RHB ha^-1 have a stronger relationship with the CCSi and ASi. Huang et al. (2020) found that PASi concentration was positively correlated with plant phytolith parameters (F = 8.1, P = 0.001), particularly in Si fertilizer-amended soil subtropical bamboo in China. Tan et al. (2021) reported that soil nutrient availability significantly influenced the spatial variability of phytoliths and PhytOC concentration in rice. Since Si is the most abundant element in phytoliths, and PASi in soil has a strong correlation with straw phytolith concentration, soil nutrient availability may influence phytolith concentration by affecting plant-available Si concentration. There was a significant correlation among the SOC, phytolith and PhytOC content in different organs of rice. Similarly, Yang et al. (2020) reported a significant positive correlation between CCSi and ASi (soil phytolith) when phytolith-rich rice residue was applied. This could be due to the accumulation of soil organic matter, which when mineralized, releases phytoliths. As a result, SOC may affect the phytolith concentration by influencing soil Si availability. Contrary to our findings, Tan et al. (2021) and Ji et al. (2019) reported that SOC decreased with the increasing level of Si input. According to some studies, soil silica availability is significantly positively correlated with soil pH (Schaller et al., 2019; Majumdar and Prakash, 2021). As a result, soil pH may have a positive correlation with phytolith concentration, which is consistent with our findings. N, P, K, Si fertilization and biochar addition not only increase biomass production but also increase phytolith production and PhytOC sequestration (Song et al., 2017). This study on the effect of different Si sources (DE, SA and RHB) in rice is the first field study and a step forward from previous laboratory experiments (Guo et al., 2015; Sun et al., 2019; Li et al., 2020a). It has been previously reported that the phytolith and C content in the phytolith influences the plant PhytOC (Sun et al., 2019; Li et al., 2020a; Anjum and Prakash, 2021). The addition of different Si sources (Si fertilizers and RHB) significantly impacts the former, with increased soil CCSi and ASi as a result of Si fertilizer and RHB application.