Introduction

Over the years, nanotechnology has consistently made significant contributions and due to its precision, it combined expertise in environmental remediation1. Currently, several physical and chemical methods are available for the synthesis of nanoparticles (NPs). However, their widespread implementation is affected by their production cost2. NPs are small particles, typically ranging in size from 1 to 100 nm3. The extraordinary physicochemical properties of NPs continue to amaze researchers. In the realm of synthesising NPs, it is essential to explore highly efficient methods with rapid reaction times, despite the availability of established methodologies. Thus, it should be affordable to achieve NPs with multi-efficient and eco-friendly applications. The bio-based synthesis process offers advantages, as it is cost-effective and does not rely on harmful chemicals4. However, bio active compounds such as lipids, carbohydrates, enzymes, amino acids, and secondary metabolites play a crucial role as powerful reducing and stabilising agents, which are primarily involved in the conversion of Ag+ to Ag5.

Additionally, microalgal and cyanobacterial cultures are more easily cultivated in the laboratory as well as the industrial scale6. Cyanobacteria, in particular, have got considerable attention for their ability to synthesise nanomaterials (NMs) due to their rapid growth rate, scalability, cost-effectiveness, and their abundance in nature. Various species from the orders like Nostocales7, Oscillatoriales and Synechococcales8 have been extensively used for NMs synthesis. In a study, Lengke et al.9 successfully synthesised intracellular and extracellular AgNPs using Plectonema boryanum UTEX 485 for the first time. Another study reported that the use of Spirulina platensis and Streptomyces glaucus for the extracellular synthesis of AgNPs10. In another study, Scenedesmus abundance and Chlorella pyrenoidosa are used for AgNPs synthesis using cell extracts11. Similarly,  Sargassum muticum12, Gracilaria corticate13, Sargassum longifolium14, Acanthophora spicifera15, and Ulva Lactuca were also used to synthesise AgNPs extracellularly.

Several strategies have been employed to synthesise AgNPs more efficiently and improve their yield. The conventional method such as one factor at a time (OFAT) is extensively used for the synthesis of AgNPs. A study was conducted using the OFAT method to synthesise the green nanoflower and utilise it for the remediation of p-nitrophenol and methylene blue dye16. However, the OFAT approach have some limitations to explain the complex interactions between variables and limiting its ability to predict significant interactions between them. Therefore, the statistical approach is used for process optimisation, and it predicts good interactions among variables. Utilising the face-centred central composite design (FCCCD) within the response surface methodology (RSM) offers an effective strategy for optimising intricate processes. This involves exploring the paired, curvilinear, and linear effects of these factors on the outcome17. Thus, Javier and Camacho18 carried out experiments aimed at refining the AgNPs synthesis by leveraging the capabilities of the RSM optimisation tool. But, the RSM optimization tool still fail to explain the nonlinear effects and not consistent with multivariant analysis. Therefore, the integration of artificial neural networks (ANN) models with RSM tools has gained widespread popularity, and they excel at handling nonlinear characteristics . Thus, the hybrid machine learning tool, based on integration of genetic algorithm (GA) and artificial neural network (ANN) which proposed high-fidelity best optimal solution after training, testing and validation of datasets19.

The potential applications of nanotechnology are vast and it includes the development of nanomaterials for diverse applications 20,21. The AgNPs have been utilised more extensively as a catalytic agent for degrading harmful dyes such as methylene blue, Congo red, methyl orange, phloxine B, auramine O, p-nitrophenol and rhodamine B dye16,22,23. Previously, some reports demonstrated that green-synthesised AgNPs can achieve dye degradation rates of 90% to 100%, establishing them as highly effective agents for dye remediation24.

This study reported that the process optimization for phyco-synthesis of AgNPs from fast growing Synechococcus sp. PCC 11901 biomass and culture-free supernatant (CFS) of Chlorella sorokiniana MSP1 using the hybrid machine learning tools. The phyco-synthesised AgNPs from Synechococcus sp. PCC 11901 biomass extract and Chlorella sorokiniana MSP1 supernatant were designated as SSB-SN and CSS-SN, respectively. The biosynthesised AgNPs were further characterised through fourier-transform infrared spectroscopy (FTIR), x-ray diffraction (XRD), field-emission scanning electron microscopy-energy-dispersive x-ray spectroscopy (FESEM-EDX), transmission electron microscopy (TEM), dynamic light scattering (DLS), and thermogravimetric analysis (TGA) techniques. Additionally, the AgNPs were subjected to the degradation of hazardous dyes, such as Sudan Black and Orange-II, and the ideal degradation kinetics were evaluated.

Results

Single-factor optimisation for AgNPs biosynthesis

The bioreduction process is a key process, where AgNO3 are reduced and forms AgNPs. In this process, the aqueous solution turned darker brown when AgNO3 reacts with the bio-extract of Synechococcus sp. PCC 11901 and Chlorella sorokiniana MSP1. The phyco-synthesis of AgNPs from both bioextract were screened in the presence of light and dark conditions. The light condition was found to be most effective conditions for the phyco-synthesis of AgNPs from both bioextracts (Supplementary Figs. 1 and 2). Thus, the colour change from a milky hue to black and brown was observed visually after 9 min of incubation as compared to control. The maximum SSB-SN and CSS-SN nanoparticles production were observed at 6 mM and 2 mM concentrations of AgNO3 after reacting with the bio-extract of Synechococcus sp. PCC 11901 and Chlorella sorokiniana MSP1, respectively (Supplementary Figs. 1 and 2). The volume of the bio-extract also played a crucial role in the formation of AgNPs. As depicted in supplementary Figs. 1 and 2, the bio-extract volume (3 mL) was found to be more suitable for the synthesis of maximum AgNPs from both cultures. Incubation time is another critical factor for the process optimisation to achieve maximum AgNPs synthesis. It was found that a higher peak of phyco-synthesised AgNPs occurred at 10 h for the SSB-SN and 12 h for the CSS-SN during the reaction (Supplementary Figs. 1 and 2).

Optimisation of AgNPs production using an FCCCD model

The statistically based FCCCD approach was used to further optimise process parameters for the phyco-synthesis of AgNPs from both reported Synechococcus sp. PCC 11901 and Chlorella sorokiniana MSP1. The obtained matrix, comprising 20 experimental sets, was analysed to understand the interaction effects among all three parameters. The quadratic model was found to be the best fit model for the phyco-synthesis of both nanoparticles. The observed responses (yield %) of the experimental setups are given in the supplementary Table 1, with ranges of 9.64–54.45% for SSB-SN and 3.01–44.58% for the CSS-SN nanoparticles. As shown in the supplementary Table 1, the maximum yield was observed at input parameters such as bio-extract (2.49 mL) and AgNO3 concentration (2.42 mM), and incubation time (19.37 h) for both AgNPs. The ANOVA study reported a calculated “F” value of 54.71 and 77.12 for SSB-SN and CSS-SN nanoparticles, respectively. The obtained p-value (< 0.0001) of the model indicates its suitability for both AgNPs (Table 1).

Table 1 ANOVA of response variables in a central composite design for SSB-SN and CSS-SN.
Full size table

Furthermore, the model’s accuracy was assessed using the coefficient of determination (R2), and it was found to be 0.91 and 0.98 for the SSB-SN and CSS-SN nanoparticles, respectively. This indicates model accuracy of 91% and 98%, respectively. Additionally, the adjusted R2 and predicted R2 of the model are given in the Table 1. Moreover, the obtained values of the predicted R2 suggested a strong correlation between the input parameters and output response. The lack of fit is another strategy to evaluate the model’s accuracy. Therefore, the lack of fit in the current study was reported with p-values of 0.0879 and 0.9652 for SSB-SN and CSS-SN nanoparticles, respectively. The quadratic equation below illustrates how the model’s process variables significantly impact the output response.

$${\text{SSB}} - {\text{SN Yield }}\left( \% \right)\,=\,+\,{\text{28}}.{\text{51}}\,+\,{\text{11}}.{\text{8}}0*{\text{A}}\,+\,0.{\text{1889}}*{\text{B}}--{\text{3}}.{\text{62}}*{\text{C}}$$
$$\begin{aligned} CSS - SN{\text{ }}Yield{\text{ }}\left( \% \right)\,=\, & +\,13.96\,+\,9.44*A--6.74*B--3.52*C--4.18*AB \\ & --0.5675*AC--0.5200*BC\,+\,3.66*{A^2}\, \\ & +\,4.20*{B^2}--2.10*{C^2} \\ \end{aligned}$$

Where A= Extract/CFC volume, B= AgNO3 concentration and C= Incubation time.

The parity plot and 3D surface plot, illustrating significant interactions among the variables for AgNPs phyco-synthesis, are shown in Supplementary Figs. 3 and 4.

Artificial neural network modelling and genetic algorithm optimisation

The ANN model was developed using 20 different sets of input variables, including A (bio-extract), B (AgNO3 concentration), and C (incubation time) and one output response as AgNPs yield. This model was successively developed by using the Levenberg-Marquardt (LM) back-propagation algorithm. Fig. 1a shows the training, testing, validation, and overall correlation coefficients of the ANN model developed with significant values of 0.99, 0.91, 0.99, and 0.97 for SSB-SN. Whereas, the training, testing, validation and the overall correlation coefficient for CSS-SN were found to be 0.99, 0.99, 0.95 and 0.98, respectively (Fig. 1b). The mean square error (MSE) for SSB-SN was obtained as 3.59 at epoch 3 (Fig. 2a), while 21.44 for CSS-SN at epoch 2 (Fig. 2b). As illustrated in Fig. 2c and d, a significant outcome of the gradient, Mu, and validation checks were obtained for SSB-SN and CSS-SN AgNPs at epoch 5. The complete range of errors is divided into 20 smaller segments (defined as vertical bars), where the y-axis represents the number of samples from our dataset assigned to each segment. The zero-error line corresponds with the zero-error point on the x-axis. Fig. 2e and f represent the obtained error histogram for SSB-SN and CSS-SN, respectively.

Fig. 1
Fig. 1The alternative text for this image may have been generated using AI.
Full size image

Regression plot obtained from ANN modelling during process parameters optimization for maximum (a) SSB-SN and (b) CSS-SN nanoparticles production.

Fig. 2
Fig. 2The alternative text for this image may have been generated using AI.
Full size image

Best performance validation plot for (a) SSB-SN and (b) CSS-SN. Gradient and validation check plots of the ANN model for (c) SSB-SN and (d) CSS-SN. Error histogram of the developed model for (e) SSB-SN and (f) CSS-SN.

Furthermore, the obtained ANN model was fitted into the GA to achieve the best optimal solutions for SSB-SN and CSS-SN AgNPs synthesis, as shown in Fig. 3a and b. The other attributed parameters like current best individual for three input factors, fitness of each individual, selection function of individuals and score histogram as shown in supplementary Fig. 5. The score histogram analysis usually performed to visualize the distribution of fitness values scores attained by different individuals and helps in the understand the performance and convergence of the GA evolutionary process. In the present study, a narrow and higher peak was found to be nearly 57.134 and 46.2849 score range for SSB-SN and CSS-SN, respectively, which indicates better convergence towards the optimal solutions (Supplementary Fig. 5). These outcomes were supported and validated by a study conducted by Rufina and her colleagues25.

Fig. 3
Fig. 3The alternative text for this image may have been generated using AI.
Full size image

Best fitness values obtained from genetic algorithm optimization and transmission electron microscopy analysis of (a and c) SSB-SN and (b and d) CSS-SN.

Characterisation of SSB-SN and CSS-SN

Zeta potential

The DLS analysis of the phyco-synthesised SSB-SN and CSS-SN nanoparticles was analysed for the zeta potential (ZP) and zeta size, as shown in supplementary Fig. 6. The ZP of the SSB-SN and CSS-SN was found to be -20.95 mV and  -31.17 mV, respectively. In contrast, polydispersity index (PDI) with value of 0.22 and 0.37 were obtained for the SSB-SN and CSS-SN nanoparticles, respectively. These results suggest that the AgNPs demonstrate enhanced stability within their colloidal suspension, highlighting the degree of repulsion among charged particles in the dispersion.

Transmission electron microscopy analysis

The morphology of the phyco-synthesised AgNPs was examined using TEM microscopy (TECNAI 20G2, Thermo Fisher). The TEM analysis showed the synthesized SSB-SN had an average particle size of 10.66 nm, with a distribution range from 5.13 to 18.13 nm (Fig. 3c). In contrast, the CSS-SN nanoparticles had a larger average size of 26.03 nm and a distribution range of 19.85 to 30.38 nm (Fig. 3d) with exhibiting a spherical morphology.

Structural and elemental composition analysis

The morphological structure of both synthesised AgNPs was accessed using the FE-SEM technique (GEMINI 560, Carl Zeiss). The FE-SEM study of SSB-SN nanoparticles provided clear insights into the variable sizes with distinct cubical shapes (Fig. 4a). Whereas, the FE-SEM image of CSS-SN revealed a nearly spherical structure with a smooth surface (Fig. 4b). It also displayed multiple clusters surrounded by biomolecule layers, which contributed significantly to Ag+ reduction and stabilization of resultant AgNPs. The energy dispersive x-ray spectrometry (EDS) analysis confirmed the presence of free Ag+ ions on the surface of both SSB-SN (Fig. 4c) and CSS-SN (Fig. 4d) nanoparticles. The EDS analysis revealed a distinct signal for silver, exhibiting the highest peak intensity percentage, which confirms the successful biosynthesis. Additionally, the weight and atomic percentage revealed the successful phyco-synthesis of SSB-SN and CSS-SN, confirming the presence of silver nanoparticles (see the inset tables in Fig. 4). The sample analysis revealed the presence of stabilisers featuring alkyl chains, as confirmed by the detection of oxygen and carbon spots. The other elemental compositions associated with SSB-SN and CSS-SN synthesis are identified through EDS-mapping, as shown in supplementary Fig. 7 for SSB-SN and supplementary Fig. 8 for CSS-SN.

Fig. 4
Fig. 4The alternative text for this image may have been generated using AI.
Full size image

Surface morphology and elemental analysis of phyco-synthesised (a and c) SSB-SN and (b and d) CSS-SN using FESEM-EDX.

Functional groups analysis using FTIR analysis

The FTIR spectra of SSB-SN and CSS-SN shown in Fig. 5a confirmed the existence of functional groups associated with the AgNPs synthesis. The prominent peak at around 3388 cm− 1 indicated the stretching and vibrations of hydroxyl groups derived from the phenolic, flavonoid, and triterpenoid compounds present in the biomass of the cyanobacterium Synechococcus sp. Notably, the peak at 2927 cm− 1 corresponds to broadened (C–H) bands, affirming the existence of aliphatic compounds that play an essential role in the bio-reduction process. The peak at 3472 cm− 1 is associated with O-H stretching, indicating the presence of alcohols or phenolic compounds. The 1547 cm− 1 peak corresponds to C-O stretching, while the 1385 cm− 1 peak reflects N-O stretching vibrations, signifying nanoparticle formation. The 1028 cm− 1 peak indicates C-N stretching vibrations, pointing to aliphatic amine groups, and the peak at 714 cm− 1 relates to N-H stretching vibrations, indicating the presence of amine groups.

X-ray diffraction analysis

The XRD pattern of SSB-SN and CSS-SN nanoparticles shown in Fig. 5b confirmed their crystalline structure. The obtained pattern of the AgNPs showed that the peaks agreed with the standard data (JCPDS no. 04-0483), which reveals the crystalline nature of the phyco-synthesised SSB-SN and CSS-SN. The diffractometer displayed prominent outcomes at various diffraction angles, as summarised in Table 2. The average size of phyco-synthesised AgNPs was calculated using the Debye-Scherrer equation and found to be 26.81 nm and 27.76 nm for SSB-SN and CSS-SN, respectively.

Fig. 5
Fig. 5The alternative text for this image may have been generated using AI.
Full size image

Functional group analysis (a), crystallinity characterization analysis (b) and thermal stability analysis of SSB-SN (c) and (d) CSS-SN.

Table 2 Summary of XRD analysis of phyco-synthesised SSB-SN and CSS-SN at various diffraction angles.
Full size table

Thermal stability analysis

In this study, TGA was employed to evaluate the thermal stability of phyco-synthesised SSB-SN and CSS-SN using the Mettler Toledo DSC/TGA 1. These experiments were conducted in a nitrogen atmosphere within a temperature range of 25 °C to 800 °C, maintaining a uniform heating rate of 20 °C/min. The TGA curves and weight loss of the AgNPs are illustrated in Fig. 5c, d. Notably, there was minimal weight loss observed (approximately 1%) at 175 °C for SSB-SN and at 200 °C for CSS-SN, which is likely due to water vaporisation. Approximately, 34.16% and 33.3% weight loss was observed between 174 °C and 750 °C for SSB-SN and CSS-SN (Table 3). The substantial weight loss may be associated with the loss of capping agents, such as proteins from the extract. These outcomes are aligned with findings reported by Karageorgou et al. in 2022, in which around 4% mass loss below 120 °C due to the vaporisation of absorbed water was reported. Additionally, they also concluded that a 66% weight loss was observed in the temperature range of 120–500 °C, resulting from the decomposition of the organic capping agents26. A brief overview of the synthesis and characterization of AgNPs with their significance is presented in Table 4.

Table 3 Kinetic models and their parameters study for O-II and SB dye sorption by the SSB-SN and CSS-SN nanoparticles (O-II: Orange-II; SB: Sudan black).
Full size table
Table 4 A comparative overview of synthesis and characterization of silver nanoparticles with their significance.
Full size table

Determination of dye degradation

Effect of sorbent dose

This study examined the impact of the SSB-SN and CSS-SN dose (ranging from 20 to 500 µL mL− 1) on the removal of O-II and SB dyes. As depicted in Fig. 6a, the maximum O-II dye removal was obtained at a sorbent dose of 500 µl mL− 1 of both SSB-SN and CSS-SN. In comparison, 80 µl mL− 1 (SSB-SN) and 60 µl mL− 1 (CSS-SN) of sorbent dose were optimised for SB dye removal (Fig. 6b). At the optimised conditions, around 55.83 ± 3.28% and 42.75 ± 2.49% of O-II dye (100 µg mL− 1) removal was achieved by SSB-SN and CSS-SN nanoparticles, respectively. In the case of SB dye (100 µg mL− 1), SSB-SN resulted in an 88.02 ± 0.15% while CSS-SN demonstrated an 85.90 ± 0.25% dye removal. Our study reported that both nanoparticles are efficient in removing of SB dye as compared to O-II dye. As shown in supplementary Fig. 9, clear and significant reduction in the band intensity at 484 nm and 598 nm was observed for both O-II and SB dyes degradation in the presence of SSB-SN and CSS-SN nanoparticles. This difference is clearly because of the distinct chemical and physical properties of dyes. It was noted that O-II is soluble in water and forms stable solutions at room temperature, carrying a negative charge. Also, it was concluded that a higher sorbent dose is required for the removal of O-II dye. The possible mechanism is that the presence of sulfonic acid groups which may contribute to the stability and resistance to degradation of O-II.

Fig. 6
Fig. 6The alternative text for this image may have been generated using AI.
Full size image

Determination of % dye removal efficiency of SSB-SN and CSS-SN AgNPs. Effect of sorbent dose on (a) Orange-II and (b) Sudan black dyes. Effect of initial dye concentration of (c) O-II and (d) Sudan black dyes.

Effect of initial dye concentrations

To determine the efficiency of phyco-synthesised AgNPs, various concentrations (ranging from 20 to 500 µg mL− 1) of both dyes were treated with an optimised sorbent dose at optimal conditions. The optimum initial dye concentration was found to be 40 µg mL− 1 and 200 µg mL− 1 for O-II and SB dye, respectively. At optimal conditions, approximately 99.79 ± 1.18% and 73.13 ± 0.82% of O-II dye was removed after being treated with SSB-SN and CSS-SN, respectively (Fig. 6c). In the case of SB dye, around 98.17 ± 0.07% and 97.76 ± 0.08% of dye removal was achieved after being treated with SSB-SN and CSS-SN nanoparticles, respectively (Fig. 6d).

Contact time and pH points of zero charge

Based on the adsorbent dose analysis, dosages with 80 and 500 µL mL− 1 of SSB-SN were reported for highest SB and O-II dyes removal. whereas  60 and 500 µL mL− 1 dose of CSS-SN. The contact time ranges from 10 to 180 min were evaluated for the dyes removing efficiency at these optimized conditions. The SSB-SN and CSS-SN takes 150 and 110 min to remove O-II dyes (Fig. 7a), whereas its required only 40 and 150 min to remove the SB dye (Fig. 7b). In the current study, the pH of both SSB-SN and CSS-SN was found to be 7; however, they differ in their pHpzc with 6.77 for SSB-SN (Fig. 7c) and 6.57 for CSS-SN (Fig. 7d).

Fig. 7
Fig. 7The alternative text for this image may have been generated using AI.
Full size image

Effect of contact time on the % dye removal of (a) Orange-II and (b) Sudan black dyes. pH Points of zero charge analysis of phyco-synthesized (c) SSB-SN and (d) CSS-SN nanoparticles.

Mechanism involved in sorption kinetics

The kinetic models, such as the pseudo-first-order, pseudo-second-order, intraparticle diffusion model, liquid film model, and Elovich model, were employed to elucidate the sorption kinetics mechanism. Table 3 presents the summarised datasets of the various kinetics studies and their corresponding parameters for the sorption process. The regression coefficient values (R2) across these models indicated the model’s significance. As shown in Fig. 8; Table 3, the pseudo-second-order was found to be the most significant model, with the highest R2 value of nearly 0.999 for both AgNPs. The intraparticle diffusion model showed significant R2 values of 0.970 and 0.901 for O-II dye when treated with SSB-SN and CSS-SN nanoparticles, respectively. Conversely, the intraparticle diffusion model during SB dye degradation by both SSB-SN and CSS-SN showed the lowest R2 values of 0.381 and 0.581, respectively (Table 3). The results indicated that the adsorption of SB and O-II dyes closely followed pseudo-second-order , highlighting the adsorption at the surface of biosynthesised AgNPs. However, the removal of O-II dye by SSB-SN exhibited behaviour consistent with the intraparticle diffusion model.

Fig. 8
Fig. 8The alternative text for this image may have been generated using AI.
Full size image

Various kinetic models’ study for orange-II and Sudan black dyes degradation by SSB-SN and CSS-SN nanoparticles. (a) and (b) represent the pseudo-first order kinetic study, whereas (c) and (d) are the pseudo-second order kinetic of orange-II and Sudan black dyes, respectively. (e) and (f) showing the Elovich kinetics model for orange-II and Sudan black dyes.

Discussion

NPs have unique disparities that set them apart from their larger counterparts in terms of physical, biological, and chemical properties27. The sustainable production of nanoparticles using natural agents has emerged as a promising solution to these challenges. A diverse array of biological organisms, including lichen, bacteria, fungi, actinomycetes, and macro- and microalgae, have demonstrated an impressive ability to reduce metal precursors into nanoparticles28. Biological extracts play a crucial role in extracellular and intracellular greener synthesis processes by transforming bulk materials into nanoscale forms. Harnessing the power of bioactive substances such as pigments, phenolics, enzymes, proteins, alkaloids, and polysaccharides would significantly reduce and stabilise the nanoparticles29.

In the current scenario, there is a growing interest in synthesising nanoparticles from cyanobacteria, which is expected to accelerate research in this area. However, a variety of cyanobacteria, such as Lyngbya majuscule CUH/AI/MW-150, Dolichospermum sp., Plectonema organum UTEX 485, Anabaena sp. PCC 7120, Phormidium, Nostoc, Spirulina, and Oscillatoria have been reported for the synthesis of AgNPs30,31. It has been observed that extracellular manufacturing involves utilising the biological substances, including proteins, pigments, hormones, antioxidants, ions, and enzymes32. Apart from cyanobacteria, microalgae have also advancment in synthesising of AgNPs because they can produce various compounds that act as reducing agents, such as amines, proteins, alkaloids, pigments, carbonyl groups, flavonoids, terpenoids, and phenolics33. Various other interactions among polyphosphates, polysaccharides, and carboxyl groups on cell surfaces have been captured, resulting in the reduction of metal ions and the formation of nanoparticles28.

The surface plasmon resonance (SPR) peak was observed at a wavelength between 400 and 450 nm in UV-vis spectroscopy, confirming the synthesis of AgNPs. A study reported that a SPR peak between 410 nm to 450 nm indicates the biosynthesis of AgNPs34. According to Senthil and colleagues, the culture filtrate from Isochrysis galbana has been used for the AgNPs synthesis35. Similarly, another study reported that the synthesis of AgNPs using the bio-extract of Oscillatoria sp. and Oscillatoria willie NTDM0136. A similar type of study was conducted and reported by Rajkumar et al. in 2021, who observed the formation of AgNPsat a 3 mM AgNO3 concentration and 8:2 ratio of the bio-extract of Chlorella vulgaris33. Another study reported that the synthesis of AgNPs with algal extract37. Moreover, the development of AgNPs are further depends upon the reaction time. Therefore, in the present study, the maximum AgNPs were observed at 12 h of reaction time. Previously, a study reported and concluded that the initiation of AgNPs synthesis was initially slow but change in colour intensity recorded after 24 h of reaction time which indicates the formation of NPs33.

Furthermore, the process optimization for the maximum yield of phyco-synthesised AgNPs were carried out using the statistical tool integrated with AI based optimization modelling. As illustrated in Table 1, the ANOVA study revealed the model’s fitness and suitability with a very good correlation between the variables. A significant f-value and p-value (< 0.0001) is desirable, which predict the model significancy. Additionally, an efficient R2 value was reported with 0.91 and 0.98 for SSB-SN and CSS-SN which indicates model accuracy of 91% and 98%, respectively. He et al.,38 conducted a Box-Behnken design for the parameter optimisation of AgNPs synthesis using the cyanobacterial extract of Microcystis aeruginosa. Similar outputs are also observed in the process optimization using statistical optimization tool39. An integrated AI tool, such as GA-ANN, was further used to standardise the most suitable reaction components for AgNPs synthesis. The GA-ANN optimization approachis a hybrid heuristic tool, which proposed high-fidelity with best optimal solution after training, testing and validation of datasets. Moreover, the ANN provides the search space and utilizes GA to find the best solution by tuning the weights and biases required to achieve lower error rates. Therefore, the GA optimisation tool is proficiently optimised and delivers an optimum solution for the targeted objective. To achieve this process, firstly, a fitness file was created based on the developed ANN model, and then it was integrated with the GA tool. This fitness file is trained multiple times to achieve the optimal solution for AgNPs synthesis. A similar algorithm was used to create the ANN model for predicting the size of silver nanoparticles25. The error histogram illustrates the discrepancies between the target values and the expected outcomes after training a feed-forward neural network. These error values may be negative and represent the differences from the desired results40. A higher validation performance indicates the model has been successfully trained and evaluated to minimise the mean square error, leading to the best possible results.

In addition, the obtain model predicted the highest SSB-SN yield at optimal parameters of bio-extract (2.49 mL), AgNO3 concentration (2.42 mM), and incubation time (22.78 h). Whereas 2.49 mL of bio-extract, 8.07 mM of AgNO3 and 19.37 h of incubation time were optimised for the CSS-SN nanoparticles. At the given parameters, close values of the output response were obtained by performing the experiments in triplicate. By comparing the RSM and GA-ANN optimisation processes, the GA-ANN model predicted the most suitable and highly efficient process parameters for both AgNPs synthesis. A significant correlation coefficient with a value of 0.97 (for SSB-SN) and 0.98 (for CSS-SN) was achieved after GA-ANN optimisation. The results mentioned above indicate that the GA-ANN has better optimisation adaptability for predicting synthesis process parameters than the RSM tool. Several researchers have conducted a comparative analysis of RSM, ANN, and GA parametric optimisation for bioprocesses. Ghorbanizamani conducted a combinatorial approach incorporating an ANN study for an AgNPs-based sensor for the detection of chicken meat spoilage41. Another study reported that the utilisation of ANN modelling for the application of silver bio-nanoparticles and understanding their mechanism42.

The scalability of phyco-synthesis of AgNPs basically depends upon the various factors such as reagent concentration, reducing agents, temperature, initial pH, reaction time, and synthesis method. Green synthesis offers sustainable approaches for the large-scale production as compared to conventional method27. The advancements in green synthesis using microorganism’s extract and continuous flow system are enabling more efficient, ecofriendly and high-yield production. Currently, various synthesis techniques can be utilized and categorized in two fundamental approaches for the AgNPs synthesis: bottom-down and top-down. Many researchers reported the top-down approach offers high, pure and homogenous particle sizes of AgNPs. Although this approach does not use any potentially hazardous chemicals or stabilizing agents to stop agglomeration, it does necessitate a large amount of external energy and advanced equipment43. Chemically synthesized NPs can be applied in a variety of chemical combinations and manufactured quickly NPs. However, the possible issues connected to the synthesis process using hazardous compounds may limit their use in medicinal applications. Microbe-derived bioactive compounds like flavonoids, alkaloids, terpenoids, and phenolic compounds, enzymes, and exopolysaccharides, are used in biological synthesis, which has gain more attention over the past few decades because of its ecological benefits. However, when scaling up an AgNPs synthesis technique, special consideration must be given to the following aspects such as ensuring uniform particle size, controlling reaction conditions, managing hazardous by-products, and overcoming potential agglomeration issues44. However, process optimization is required to improve the AgNPs synthesis scalability. Thus, AI-based tools play a significant role in the process parameters optimisation to standardise the best optimal components and provide the most suitable solutions with higher accuracy. As a result, producing them on a large scale is extremely challenging, even if they are easily manufactured on a small scale. The eco-friendly synthesis of metal nanoparticles and their oxides from microbial bioextract may be a successful, economical, and sustainable substitute over the traditional techniques45.

The phyco-synthesised AgNPs were further characterized using the TEM (TECNAI 20G2. Thermo Fisher), FTIR (Spectrum Two, PerkinElmer), FESEM-EDX (GEMINI 560, Carl Zeiss), XRD (Empyrean, Malvern Panalytical) and TGA analysis (Mettler Toledo DSC/TGA 1). The ZP is an essential parameter that reflects the stability of AgNPs in an aqueous environment and their morphological characteristics. Generally, a ZP value of -30 mV often indicates a stable colloidal solution due to a balance of electrostatic forces among the particles46. Consequently, our results revealed that the phyco-synthesised AgNPs exhibit both polydispersity and long-term stability, attributed to their negatively charged surfaces, which are essential for various biological applications. Interestingly, the particle size observed in our research closely aligns with the findings of a study that reports a particle size of 10–20 nm, exhibiting a spherical shape in Chlorella sp. Similarly, TEM analysis revealed that the nanoparticle synthesizes from Synechococcus moorigangae shows a distribution ranges from 2.5 to 19.9 nm, displaying a spherical shape with an average particle size of 8.2 nm47. In another study, biosynthesised AgNPs from Chlorella sp. MF1 were spherical, and measuring size between 1.45 to 5.08 nm48. Furthermore, the biosynthesised nanoparticles demonstrated excellent dispersion and were free from substantial agglomeration or variations in their structure, ensuring optimal performance and reliability.

Furthermore, the FTIR results of the phyco-synthesised AgNPs has shown a variety of the functional groups, which indicates the presence of bioactive molecules and play a crucial role in the stabilizing and capping the AgNPs synthesis49. Similar stretching and vibrational analysis were observed by Arévalo-Gallegos and their colleagues50. TheAgNPs pattern showed peaks that agreed with the standard data (JCPDS no. 04-0783), indicating the crystalline nature of the phyco-synthesised SSB-SN and CSS-SN51. The present finding consistent with study conducted and reported by Sidorowicz and colleagues52. The result of the TGA curve of the AgNPs has proved that the SSB-SN and CSS-SN nanoparticles is in highly stable at higher temperature by losing only 34.16 and 33.3% weight loss. These outcomes are better than the findings reported by Karageorgou et al.26.

The eco-friendly degradation of O-II and SB dye was observed by using the phyco-synthesised AgNPs. Our study reported that both nanoparticles are highly efficient in removing SB dye compared to O-II dye. Also, it was concluded that a higher sorbent dose is required for the removal of O-II dye. The possible mechanism is that the presence of sulfonic acid groups may contribute to the stability and resistance to degradation of O-II. A study found that the maximum removal efficiency of methylene blue was achieved at an initial dye concentration of 100 µg mL− 1, with Chlorella vulgaris reaching a removal rate of 83.04%53. Similarly, Rajkumar and his colleague’s synthesizes the AgNPs from the C. vulgaris and found 96.51% of methylene blue dye degradation within 3 h of incubation time33. Thus, the initial dye concentration has a significant impact on the effectiveness of dye removal ifficiency. A similar study conducted by using the Synechococcus sp. cell free extract synthesised AgNPs, which have efficient dye degradation ability54. According to Lanjwani and co-workers, around 90% of Sudan black dye was degraded using SnO2 nanoparticles as catalyst55. Another study reported that gold nanoparticles degraded around 90% and 25% of methyl orange and Sudan dyes, respectively56. In addition, chemically synthesised silver nanoparticles (CH-AgNPs) have similar efficiency to degrade the orange and blue dyes with 97.4% and 100% in the presence of NaBH457. Rodríguez-Cabo and co-workers reported that around 79.2% of Orange-II dye was degraded from silver chloride nano-catalyst in the presence of UV-light58.

It is well known that, the possible mechanism of dye degradation is dependent on adsorption of dyes on the surface of nano-catalyst and release of charge generated in which holes and electrons are photochemically generated on the surface of photocatalyst. Afterward, the generated electrons transfer from valence band (VB) to conduction band (CB), creating the positively hole (h+) charged in VB59. Oxidoreduction reactions occur on the photocatalyst surface and then molecular oxygen catalysed into reactive oxygen species (ROS) such as free radical and superoxide radical (O2−) via exciting electrons60. The ˙OH free radical generated after oxidation of hydroxyl ion by the action of excited electrons. Thus, the generated hole and electrons act as primary active substances, undergo redox reactions and create ˙OH/O2− secondary active substances/free radicals58. The O2− radicals reacts with electron hole pairs and created H2O2, O˙ and hydroxyl radicals. These ROS reacts with the targeted dye molecules and breaks into smaller and nontoxic fragments via oxidation degradation. Precisely, the dyes molecular structure disrupted by the action of ˙OH radicals61.

In the current study, the pH of both SSB-SN and CSS-SN was found to be 7; however, they differ in their pHpzc, with 6.77 (SSB-SN) and 6.57 (CSS-SN), respectively. The parameter (pHpzc) indicates the extent of ionisation on the adsorbent’s surface. When the adsorbent is positively charged, it attracts anions at pH values below the pHpzc, whereas when it is negatively charged, it attracts cations at pH values above the pHpzc16. SB is typically regarded as a non-ionic dye, lacking a significant positive or negative charge under neutral to slightly acidic pH levels. This dye can interact with the charged surface of adsorbents through hydrophobic interactions and Van der Waals forces rather than through electrostatic attraction or repulsion62. However, O-II removal can still occur through hydrophobic interactions or Van der Waals forces, particularly at slightly acidic pH (4–6)63. Adjusting the pH to slightly acidic levels significantly enhances dye adsorption by decreasing the negative charge on the nanoparticles’ surfaces and strengthening the interactions. The initial pH of the reaction had a substantial impact on the adsorption process, influencing both the dye ionisation and the availability of binding sites on the adsorbent.

Various kinetic models, such as the pseudo-first-order, pseudo-second-order, intraparticle diffusion model, liquid film model, and Elovich model, were employed to elucidate the sorption kinetics mechanism, as shown in Table 3. The R2 value across these models indicated the model’s significancy. The results indicated that the adsorption of SB and O-II dyes closely conforms to the pseudo-second-order model, highlighting the adsorption rate onto the surface of biosynthesised AgNPs. Similar, finding reported by Lanjwani and their colleagues during SB dye degradation55. In a study, Kgatle and coworkers reported the pseudo-second-order kinetic model which was the most suitable for the methyl orange dye degradation using the Fe/Ag/Zn trimetallic nanoparticles64. In contrast, first-order reaction kinetic model was followed during the photodegradation of Orange-II dye with the help of silver chloride nanoparticles58. However, the degradation of O-II dye by SSB-SN exhibited behaviour consistent with the intraparticle diffusion model. An investigation reported the effectiveness of AgNPs synthesised from Anabaena variabilis and Spirulina platensis as biosorbents for removing malachite green dye, following first-order kinetics65.

Methods

Chemicals and reagents

Chemicals such as potassium dihydrogen phosphate, sodium chloride, sodium nitrate, magnesium sulphate heptahydrate, disodium EDTA, copper (II) sulfate pentahydrate, calcium chloride dihydrate, Vitamin B12, and Orange-II were purchased from Sigma-Aldrich, India. Silver nitrate (AgNO3; 99.9% purity) and Sudan black dye were purchased from SRL, India. All the chemicals used in the present study were of analytical grade.

Microorganisms’ conditions

The cyanobacteria Synechococcus sp. PCC 11901 was cultivated in the MAD medium at 250 rpm shaking, under continuous illumination, using RGB LEDs at a 1:1:1 ratio and a constant light intensity of nearly 750 µmol photons m−2s− 1. The cultivation was carried out at 30 °C with a shaking incubator (CIS-24 Plus, REMI). On the other hand, the microalgae Chlorella sorokiniana MSP1 were grown and maintained in BG-11 medium under a light intensity of 28-35 µmol photons m-2s-1 at 28 °C.

Microalgal and cyanobacterial extract preparation

The Chlorella sorokiniana MSP1 was grown in fresh BG-11 medium under light conditions of 25–40 µmol photons m−2s− 1. The culture was first initiated in 250 mL sterile flasks containing 100 mL BG-11 medium. The sample was centrifuged at 8000 rpm for 10 min to obtain the supernatant, which was then used as the bio-extract for the experiment. On the other hand, the biomass of Synechococcus sp. PCC 11901 was obtained by centrifugation at 8000 rpm for 15 min. Subsequently, the pellets obtained were gathered and dried in a petri dish for 24 h at 60 ℃. The dried biomass was then pulverised into fine powder using a mortar and pestle. The pulverised biomass powder was dissolved in distilled water and kept overnight at 30 ℃ and 120 rpm, marked as biomass extract.

Phyco-synthesis of silver nanoparticles

For the phyco-synthesis of AgNPs, a reaction mixture consisting of 3 mL of bio-extract and 7 mL of 5 mM AgNO3 solution was mixed and incubated at room temperature for 36 h. Subsequently, a brown colouration was observed, indicating the synthesis of AgNPs, and the absorbance was measured at a wavelength range of 200–800 nm using a microreader plate (FlexA-200, Genetix). The biosynthesised AgNPs from Synechococcus sp. PCC 11901 biomass extract and Chlorella sorokiniana MSP1 supernatant were designated as SSB-SN and CSS-SN, respectively.

Single-factor optimisation approach

The biosynthesis of AgNPs by Chlorella sorokiniana MSP1 and Synechococcus sp. PCC 11901 was achieved through the optimisation of a single-factor optimisation approach. This approach utilised a single-factor screening called OFAT to evaluate various physicochemical parameters, including the bio-extract volume, AgNO3 concentration, and incubation time. The impact of varying AgNO3 concentrations, ranging from 0.5 mM to 10 mM, on the biosynthesis of AgNPs was studied. We also assessed different volumes of bio-extract for the biosynthesis of AgNPs. Furthermore, various incubation times ranging from 2 to 18 h were evaluated for the biosynthesis of AgNPs.

Experimental design for process optimisation modelling

Statistical modelling was conducted to optimise the process further and achieve the maximum yield of AgNPs production. Design-expert software (DoE ver. 13.0) was used to obtain the optimisation model, which involved an FCCCD with three independent input parameters: bio-extract (A), AgNO3 concentration (B), and incubation time (C). A matrix having 20 experimental runs was obtained from the FCCCD model (Supplementary Table 1). The experimental results of FCCCD were fitted with a second-order polynomial model using the Eq. (1).

$$Y={\beta _{i0}}+\mathop \sum \limits_{{i=1}}^{3} {\beta _i}{X_i}+\mathop \sum \limits_{{i=1}}^{2} \mathop \sum \limits_{{j=i+1}}^{3} {\beta _{ij}}{X_i}{X_j}+\mathop \sum \limits_{{i=1}}^{3} {\beta _{ii}}X_{i}^{2}$$
(1)

Where Y is the AgNPs yield (%) response. β0, βi, βij, and βii are the regression coefficients for intercept, linear effects, two-way interactions, and square terms, respectively.

A comprehensive analysis of the strong correlation among the studied variables was carried out using ANOVA and a 3D surface response plot. The model accuracy was evaluated by the values of the R2 and the adjusted R2 (adj-R2). At the same time, model significance was determined by the ANOVA study.

Development of the artificial neural network model

The ANN model was further developed to predict the maximum AgNPs yield from the obtained FCCCD experimental data sets. The adopted model comprises one input layer having three independent predictors, one hidden layer and one output layer with one response, i.e., AgNPs yield. In this model, the first layer of neurons collects significant information, which is subsequently transmitted to the hidden layers. Ultimately, the result is drawn from the final layer of neurons. The relationships between the units were fine-tuned until a minimal prediction error was achieved across various training algorithms; evidently, the system reached that specific value38. The FCCCD established a design matrix involving three input parameters: A (Bio-extract), B (AgNO3 concentration), and C (Incubation time), which were used for the process optimisation using ANN. This setup illustrates a nonlinear relationship and is used to train, test, and validate the model. During the forward propagation phase, errors are computed by multiplying all inputs by arbitrary weights (derived from the FCCCD experimental data), followed by the application of the tangent sigmoid activation function. The weights are updated using the gradient descent method based on the errors calculated by the feed-forward back-propagation network (FFBP) during the backwards propagation process. The efficiency and reliability of the Levenberg-Marquardt (LM) backpropagation algorithm have been recognised and utilised for this purpose25.

Genetic algorithm-artificial neural network (GA-ANN) optimisation

The developed ANN model was integrated into a Genetic Algorithm (GA) fitness function. This integration aims to optimise the production of AgNPs. Key components of GA include chromosome length, mutation, crossover and fitness selection, where progeny demonstrating the highest output of AgNPs are favoured, as well as various operators that influence AgNPs production. Initially, the GA selects a strong solution from a population set. As this population evolves, new progeny is generated, and through natural selection, the more successful offspring are retained, while the less favourable ones are filtered out. The following Eq. (2) was used to optimise the process parameters for the production of AgNPs.

$$~l={N_h}+\left( {{N_i} \times {N_h}} \right)+\left( {{N_h} \times {N_o}} \right)+({N_h}+{N_o})$$
(2)

Where Nh represents the hidden layer neurons. While Ni and No are the number of predictors and the output response.

The parameters, such as population size and maximum generation, were set to 500 and 200, respectively, for the adaptation of GA optimisation. In contrast, other parameters were kept constant at their default settings in MATLAB 2017b. A flowchart of the process optimization for phyco-synthesis of AgNPs using the integrated hybrid tools are illustrated in the Fig. 9.

Fig. 9
Fig. 9The alternative text for this image may have been generated using AI.
Full size image

The process flow chart of face centred central composite design (FCCCD) integrated with artificial neural network (ANN)-genetic algorithm (GA) for efficient phyco-synthesis of silver nanoparticles.

Characterisation of phyco-synthesised AgNPs

The colour changes from milky to brown were measured at ranges from 200 to 800 nm. A band intensity between 400 and 450 nm was observed, indicating the formation of AgNPs. The zeta potential of the phyco-synthesised AgNPs was measured using a zeta-sizer system (ZSU5700, Malvern Panalytical, UK). The presence of functional groups associated with AgNPs biosynthesis and capping was assessed at 400–4000 cm− 1 using the Spectrum Two PerkinElmer FTIR spectrophotometer16. The size of both synthesised AgNPs was determined through TEM analysis using a Thermo Fisher TECNAI 20G27. The surface morphology and elemental composition of synthesised AgNPs were investigated using a FESEM equipped with an EDX system (GEMINI 560, Carl Zeiss)47. XRD analysis was carried out using an XRD EMPYREAN diffractometer with a Cu Kα filter at angles ranging from 20° to 70° of 2θ52. The thermal stability of biosynthesised AgNPs was analysed using the TGA technique (Mettler Toledo DSC/TGA 1). For the TGA analysis, the synthesised AgNPs were performed at a temperature range from 20℃ to 800℃ in air with a ramp of 20℃/min.

Investigation of the optimal operational parameters for dye degradation

A series of batch experiments was conducted to investigate the catalytic effectiveness of the SSB-SN and CSS-SN in effectively removing Sudan Black (SB) and Orange-II (O-II) dyes. This study investigated the primary factors, including the SSB-SN and CSS-SN dosage, contact time, and initial dye concentration, that contribute to the effective degradation of dye. A negative control was also performed without adding the AgNPs in the reaction set.

Zero charge of pH point (pHpzc)

The pHpzc for SSB-SN and CSS-SN was determined using the method described by Singh et al.66. Briefly, 20 mL of 0.1 mM NaCl solution, containing 2 mL of SSB-SN and CSS-SN each, was pre-adjusted to a pH range of 2 to 11. After 24 h of incubation, we measured the final pH and plotted the difference in pH values against the initial pH values.

Effect of sorbent dose

To investigate the effect of different dosages of SSB-SN and CSS-SN on the removal of SB and O-II, a range of sorbent concentrations – precisely 20, 40, 60, 80, 100, 200, 300, 400 and 500 µl mL− 1 were tested under standard pH, 30℃, 120 rpm and incubation time of 180 min. The reaction mixture, having each dye solution (100 µg mL− 1), was treated with different sorbent doses and incubated under specific conditions. Then, the absorbance at 598 nm for SB and 484 nm for O-II dye was measured using a FlexA-200 microplate reader (Genetix). The dye removal efficiency (R%) was calculated using the formula described in Eq. (3)16,67.

$$R\left( \% \right)=\frac{{{C_i} - {C_t}}}{{{C_i}}} \times ~100$$
(3)

where Ci and Ct are the initial and final concentrations of the dyes, respectively.

Effect of initial dye concentration

The effect of initial dye concentrations of SB and O-II was examined to achieve maximum dye degradation at an optimised sorbent dose, while maintaining other experimental conditions constant. The effectiveness of dye removal (R%) and the quantity of sorbed dye (qe, mg g− 1) were assessed using Eqs. (3) and (4), respectively.

$${q_e}=\frac{{V\left( {{C_o} - {C_e}} \right)}}{m}$$
(4)

where qe represents the amount of sorbed dyes per gram of AgNPs, V denotes the volume of solution, C0 and Ce are the initial and final dye concentrations, respectively, and m stands for the weight of AgNPs used.

Effect of contact time and sorption kinetics

The effect of contact time on maximum dye removal was investigated with some modifications to the protocol described by Mansee et al.16. Briefly, 5 mL of dye solution (optimised concentration) was prepared and mixed with appropriate SSB-SN and CSS-SN nanoparticles. Simultaneously, a control set was prepared without the addition of nanoparticles. Both experimental setups were incubated at standard parameters for up to 180 min. At different time intervals, aliquot amounts of samples were taken, and the absorbance was measured at the specific λmax for the calculation of dye removal efficiency and equilibrium.

To understand the adsorption mechanisms of orange-II and Sudan black dyes using both SSB-SN and CSS-SN nanoparticles, the adsorption mechanisms were examined by employing various kinetic models, including the pseudo-first-order, pseudo-second-order, intraparticle diffusion model, liquid film model, and Elovich model.

Conclusion

The phyco-synthesis of AgNPs using Synechococcus sp. PCC 11901 and Chlorella sorokiniana MSP1 offer a significant approach for nanoparticle production. An integrated FCCCD-ANN-GA hybrid tool was employed to optimise and predict the optimal solution for enhanced AgNPs phyco-synthesis. The obtained R² value validates the ANN-GA model’s fitness and suitability compared to the FCCCD model. Afterwards, it was revealed that the phyco-synthesised AgNPs showed smaller sizes and cubic shapes, with the presence of Ag+ ions, as analysed by TEM and FESEM-EDX techniques. The XRD and TGA analyses exhibited crystalline characteristics with well-defined interplanar spacing of phyco-synthesised AgNPs, along with considerably good thermal stability. Furthermore, the SSB-SN and CSS-SN demonstrated an efficient capability to degrade hazardous dyes e.g. SB and O-II, which follows the pseudo-second-order kinetics. Further studies are required to understand the role of phyco-synthesized AgNPs, towards the industrial dyes remediation.