Role of pH in the synthesis and growth of gold nanoparticles using L-asparagine: a combined experimental and simulation study

100 ml of 100 mm Asn solution was prepared and separated into two vials of 6 ml solution. Different ratios R = [NaOH]/[Asn] of 0.05 and 0.5, with an interval step of 0.05 were created to obtain a pH variation on each sample. The NaOH/Asn mixtures (pH-adjusted Asn solution) were left, in closed vials, under constant stirring for 3 h at room temperature and under ambient conditions to reach pH equilibrium, as shown figure 1. Then, 6 ml of the pH-adjusted Asn solution was added to 5 ml of a 1 mm fresh Gold(III) chloride acid trihydrate (HAuCl₄ ⋅ 3H₂O) solution at room temperature. The mixtures were immediately mixed and incubated at 80 °C for 40 min using a hotplate. The samples were left to cool down at room temperature and without external cooling source. Subsequently, the set of samples were centrifuged at 1000 rpm, and the supernatant was extracted as the final product. Finally, we chose two pH, one more acid (pH 6) and one more basic (pH 9), for our studies.

Zoom In Zoom Out Reset image size

Particle size, morphologies, and interplanar spacing in AuNPs were studied using a transmission electron microscope (TEM, JEOL JEM-2010) and high-resolution (HR)-TEM (Tecnai F20 microscope with FEI, acceleration voltage 200 kV) respectively. Gatan Digital Micrograph software was used to analyse the HR-TEM micrographs. Particle distribution and population were determined by ImageJ software with standard plugins, included for ∼100 particles per sample. The polydispersity (P) was estimated as

$$\begin{equation}P=\frac{\sigma }{s}100\%,\end{equation} \tag{ 1 }$$

with σ the standard deviation, σ = FWHM/2, FWHM the full width at half maximum of the total distribution, and s is the mean particle diameter (in nm).

The crystallographic toolbox (CrysTBox) software was used to interpret diffraction patterns obtained from HR-TEM micrographs [28]. The vibrational mode of Asn was characterised by Raman scattering. Raman spectra were acquired on a modular multi-channel Raman spectrograph Jobin Yvon–Spex 270M in 90° scattering geometry using a 532.15 nm line of a continuous-wave solid-state Nd-YAG laser for excitation, as described in references [29, 30]. The power at the sample was 240 mW [supplementary figure 3 in the SI (https://stacks.iop.org/CM/33/254005/mmedia)]. Ramanmeasurements were performed at room temperature-controlled hermetical quartz cell (2 mm). Spectra were acquired with 5 s accumulation time, 12 accumulations and 15 spectra per sample. Wavenumber scales were precisely calibrated (±0.1 cm⁻¹) using the emission spectra of a neon glow lamp taken before and after each Raman measurement. The Raman contribution from the water was subtracted, and the spectra were corrected for the non-Raman background.

We investigate the adsorption of Asn on (111), (100), (110) and (311) gold surfaces (figure 2) by first calculating the potential of mean force (PMF) resulting from pulling the amino acid away from the surface and towards the bulk water. Gold surfaces are oriented with normal vector parallel to the z-axis. Initially, we place the Asn as close as possible to the gold surface, at a distance of ∼0.3 nm. The simulation boxes were filled with TIP3P water molecules [31]. The centre of mass (COM) of Asn molecule was pulled out in the z direction and snapshots were collected each 0.1 nm in order to generate the starting configurations for the umbrella sampling calculation. For each window, 9 ns of MD simulation were performed with the Asn COM constrained using a harmonic potential with spring constant of 1000 kJ mol⁻¹ nm⁻².

Zoom In Zoom Out Reset image size

The PMF for each system was reconstructed using the weighted histogram analysis method [32]. All simulations were carried out with GROMACS 5.1.2 [33, 34], using CHARMM general force field for organic molecules [35, 36], for both Asn states, and a polarisable force field for gold [37].

In the experiment, different values of pH are associated to an excess in the concentration of zwitterionic (acidic, pH 6) or anionic (basic, pH 9) states of Asn. We use a major approximation to mimic these pH conditions by studying the adsorption of a single Asn molecule in two states: zwitterionic and anionic (figure 3). To evaluate the adequacy of our approximation, we perform density functional theory simulations for a single Asn molecule in the two charge states and compute their corresponding Raman spectra (supplementary figure 6(a) in the SI). As expected, the spectrum of the anionic state exhibits suppressed resonances related to vibrations of the –${\mathrm{N}\mathrm{H}}_{3}^{+}$ group. To a lesser degree, this effect is also apparent in the experimental Raman spectra of Asn in bulk solution at both pH conditions (figure 5(a)). This correspondence suggests that the simplification of identifying molecular states with pH conditions might be useful to approximate and investigate the complex experimental conditions, which are inaccessible to computer simulations that explicitly include all the chemical species present in the system.

Zoom In Zoom Out Reset image size

3.1.1. Morphological characterisation

TEM images (figure 4) show AuNPs (at 200 nm resolution) obtained by reacting gold(III) (HAuCl₄ ⋅ 3H₂O) with Asn at acidic (pH 6, figure 4(a)) and basic (pH 9, figure 4(b)) conditions. These images show that both reaction conditions yield non-spherical particles with increasingly amorphous shapes apparent at pH 9 (see also figures 3 and 4 in the SI). The average particle size was determined by analyzing twenty TEM micrographs and a minimum of 56 particles per sample (supplementary figures 1 and 2 show a gallery of TEM micrographs used in the calculation of the particle size distribution), and using standard imaging protocols with the ImageJ software package. The results from the particle size analysis are given in the histograms depicted in figure 4(c) (pH 6) and figure 4(d) (pH 9). A Gaussian fit to the size distribution shows that the pH 6 reaction conditions yield particles of 18 ± 12 nm diameter, while the pH 9 reaction condition resulted in larger particles of 43 ± 27 nm diameter. The polydispersities for each sample set were similar, calculated as 67% for pH 6 and 64% for pH 9. These results indicate that, on average, the pH 6 particles are approximately 25 nm smaller in diameter than the pH 9 particles.

Zoom In Zoom Out Reset image size

The electron diffraction pattern from the TEM measurements are shown in figures 4(e) and (f), and the corresponding electron diffraction profiles of figures 4(e) and (f) are shown in figures 4(g) and (h) for pH 6 and 9 respectively. The primary diffraction peaks correspond to (111), (200), (220), and (311) reflections, which are characteristic of the gold fcc structure [38]. The relative intensities of the diffraction peaks vary between the two sample sets, suggesting a difference in the particle shape.

The surface structure of the particles were further analyzed using high-resolution TEM (supplementary figures 3 and 4). For the pH 6 samples, two well-resolved (222)-type crystallographic planes were detected (measured D-spacing of about 1.16 Å and 1.14 Å, as shown in supplementary figure 3(c)), which run perpendicular to the sidewalls of the structure. The same process was used in the sample at pH 9, but in this case, was not apparent to identify crystalline orientations on the particle surface. Only one vertex selected shows (111)-type crystallographic planes (measured D-spacing of about 2.35 Å, as shown in supplementary figure 4(c)), which like the previous case run perpendicular to the sidewalls of the structure. The high-resolution TEM images at 10 nm showed differences in particle surface geometry between the two cases (supplementary figures 3(b) and 4(b)). In the case of pH 6, it was possible to identify four triangular partitions on the particle surface with angles of 114.5°, 87.3°, 63.1°, and 28.2° (supplementary figure 3(b)). In comparison, with the particles at pH 9, it was not possible to identify partitions on the particle surface. The pH 9 sample could not be similarly fit further indicating highly homogeneous surface structure.

The morphological characterisation indicates that the pH 6 samples are on average smaller and more homogeneous in surface geometry than the pH 9 samples.

Earlier work in the literature has shown that when Asn is used as a reducing agent to produce AuNPs, the size, color, and shape do not depend on the reactant concentrations [11]. Our results here demonstrate that when the reaction pH is used as a tuning parameter, it is possible to modify the critical properties of size, color, and shape.

3.1.2. Vibrational spectroscopic characterisation

Figure 5 shows the measured Raman spectra of the AuNPs, along with two solutions of Asn in water, at pH 6 and 9, for reference. Figure 5(a) shows the normalised Raman spectra of the reference solutions in the frequency region of 300 cm⁻¹ to 1750 cm⁻¹, along with mode assignments. The mode ν(${\mathrm{C}\mathrm{O}}_{2}^{-}$) shows the highest intensity for both solutions. In the frequency region of 700 cm⁻¹ to 1000 cm⁻¹, the Asn solution at pH 9 shows a decrease in the intensity of the modes compared to the Asn solution at pH 6 (figure 5(a)). Relevant shifts of the peaks in the spectra were not detected. Figure 5(b) shows a comparison of the normalised Raman spectra for the Asn solution in water at pH 6 and the AuNPs/Asn sample at pH 6. The sample AuNPs/Asn at pH 6 presents a peak at 400 cm⁻¹, which is associated to a torsion of –${\mathrm{N}\mathrm{H}}_{3}^{+}$ group (τ(${\mathrm{N}\mathrm{H}}_{3}^{+}$)). The mode τ(${\mathrm{N}\mathrm{H}}_{3}^{+}$) was not detected on the solution of Asn in water. In the region of 987 cm⁻¹ the stretching mode of ν(CC) underwent an increase in intensity compared to the spectrum of the solution of Asn in water. Figure 5(c) shows the comparison of the solution of Asn in water at pH 9 and the sample of AuNPs/Asn at pH 9. In this case, the spectrum of the AuNPs/Asn sample at pH 9 lacks the peak corresponding to the torsion mode of the –NH₂ group at 522 cm⁻¹. Figure 5(d) compares the Raman spectra of the samples AuNPs/Asn at pH 6 and AuNPs/Asn at pH 9. The main differences are the presence of the modes τ(${\mathrm{N}\mathrm{H}}_{3}^{+}$) and ν(CC) in the sample AuNPs/Asn at pH 6 compared to the sample at pH 9, and a peak shift towards higher energies in the region of 1125 cm⁻¹, associated with out-of-plane bending mode of γ(NH₂) group for the sample with lower pH.

Zoom In Zoom Out Reset image size

A summary of the assignment of the Raman peaks for Asn in the frequency region of 300 cm⁻¹ to 1750 cm⁻¹ is shown in table 1.

Table 1. Frequency modes of Asn and assignments in the region of 300–1750 cm⁻¹. Abbreviations indicate: s = strong, m = medium, w = weak, v = very, sh = shoulder, sc = scissoring, r = rocking, wag = wagging, tw = twisting, ν = stretching, δ = bending, γ = out-of-plane bending, τ = torsion, s = symmetric, a = antisymmetric.

Frequency (cm−1)	Assignments	Reference
364	δ(CCC); δ(CCα N); τ(CC)	[39]
400	τ(${\mathrm{N}\mathrm{H}}_{3}^{+}$)	[40]
455
522	τ(NH2)	[41]
608	δ(${\mathrm{N}\mathrm{H}}_{3}^{+}$) + δs(CONH2)	[39]
700
776
802	δr(CH2)/δtw(NH2)	[42–44]
825	γ(${\mathrm{C}\mathrm{O}}_{2}^{-}$)	[39]
836	γ(NH2)	[41]
865	r(CH2)	[40]
884	δr(CH2)	[43]
928	r(NH2)	[44]
987	ν(CC)	[42]
1077	ν(CN)	[39]
1125	γ(NH2)	[39]
1231	δtw(CH2)/δwag(CH2)	[40]
1266	w(CH2)	[40]
1329	δ(CH)	[44]
1357	δ(CH) δs(CH)	[42]
1404	δ(CN) AIII	[44]
1420	δs(${\mathrm{C}\mathrm{O}}_{2}^{-}$)/δsc(CH2)	[40]
1617	δ(${\mathrm{N}\mathrm{H}}_{2}^{+}$) amide II	[42]
1680	δ(${\mathrm{N}\mathrm{H}}_{3}^{+}$)	[44]
2129
2940	νus(CH2)	[41]
3195
3233	Water	[39]
3428	Water	[39]

We investigate the adsorption mechanism of Asn on gold surfaces with orientations corresponding to the planes (111), (100), (110) and (311), which are those that most commonly appear in the experimental TEM spectra. We compute the PMF to quantify the affinity of the Asn with these surfaces. In figure 6 we present free energy profiles resulting from pulling Asn in the zwitterionic and anionic states which we associate to the two experimental conditions, namely pH 6 and pH 9, respectively. For all the considered surfaces, it is apparent that the anionic Asn tends to bind more strongly to the gold surfaces. It is also clear that close to the surface (∼0.35 nm) the binding strength follows the trend (311) < (110) ∼ (100) < (111) for the zwitterionic Asn. For the anionic state, the trend changes to (111) ∼ (100) ⩽ (311) < (110), where the PMF for (111) gains 2.9 kJ mol⁻¹ upon increasing pH.

Zoom In Zoom Out Reset image size

To shed light on the specifics of the adsorption process, we first focus on the minimum distance between the Asn COM and the surface. These distances are distributed between 0.3 nm and 0.4 nm. In figure 7 we show a projection of the density of the water oxygens, parallel to the surface, taken over a production run of 9 ns, and at a distance between 0.30 nm and 0.60 nm. The Asn COM is constrained along the z-direction and is free to move on the surface. From figure 7, the voids present for the two Asn states and the (100)-, (110)- and (311)-oriented surfaces indicate that the Asn remains pinned on the surface and it does not diffuse during the simulated timescale. By contrast, the amino acid freely diffuses on the (111)-oriented surface. As presented in figure 8, this effect is the result of the water ordering occurring in close proximity to the gold surface, in close registry with the fcc lattice. For the (111) case, this ordering is only observed for the first water layer (at ∼0.3 nm). For the remaining surfaces, the effect is apparent up to distances between 0.6 nm and 0.9 nm. Given the polar character of the Asn, it is not surprising that it forms a rather stable hydrogen network with water molecules close to the gold surface. Furthermore, for (100), (110) and (311) surfaces where the water structuring grows into the bulk, the hydrogen network creates a pocket for the amino acid that hinders its surface diffusion. This picture remains essentially the same for the two Asn states considered here.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The rather uniform (111) surface facilitates the adsorption of Asn as can be seen in the PMF results. The amino acid lies essentially flat thus maximising its contact area with the surface. Upon pulling away the amino acid, we observe that the resulting PMF is somewhat barrierless, indicating that Asn is uniformly detaching from the (111) surface. For the other surfaces, the various minima in the PMF are associated to tilting angles between the amino acid and the surface, formed during the detaching process.

When going from the zwitterionic to the anionic state, the –${\mathrm{N}\mathrm{H}}_{3}^{+}$ group is replaced with a –NH₂ group. This change does not significantly modify the adsorption on the uniform (111) surface, as evidenced by the PMF (figure 6). For the open surfaces, this change in the charge of the amino acid results in an increase in the amino acid/surface attraction due to the presence of two –NH₂ groups which bind more favourably to gold. Figure 9 shows the distribution of distances of the amide group (black line) and the ammonium/amine (blue line) for the surface orientations considered when the COM of the Asn is at ∼0.4 nm from the surface. The wide distribution without a clear preferential binding between the two groups is apparent for the (111) case. For the remaining surface orientations at acidic pH, it is clear that the amide group has the tendency to stay closer to the surface. At basic pH, both amine and amide groups are equally spaced from the surface. The complex distributions observed for the (110) case are the result of the tendency for Asn to align with the rows present on the surface.

Zoom In Zoom Out Reset image size

These observations allows the interpretation of the Raman spectra presented in figure 5. At pH 6, panel (b), we observe that a peak at approximately 400 cm⁻¹ develops in the presence of AuNPs. This frequency corresponds to the τ(${\mathrm{N}\mathrm{H}}_{3}^{+}$) mode. Upon adsorption, the amine group, –NH₂, binds to the surface while the ammonium group, –${\mathrm{N}\mathrm{H}}_{3}^{+}$, is free to move and vibrates at its characteristic frequency, thus generating the peak in the spectra. At pH 9, panel (c), the amino acid is mostly present in its anionic form with the ammonium replaced by an amine group. As shown in the figure, the adsorption of these two –NH₂ groups on the gold surface suppress the corresponding τ(NH₂) mode at approximately 550 cm⁻¹.

The molecular orientation also provides a possible interpretation of the diffraction pattern and TEM measurements. The equilibrium shape of a metal NP corresponds to the minimum surface area configuration that minimises the surface Gibbs free energy. By combining extended Wulff construction [45] approaches with known values of surface energies, it is possible to determine the NP final equilibrium shape [46]. In the case of Au surfaces in vacuum, there is an approximately linear relation between surface energy, σ, and number of broken bonds per unit area [24]

$$\begin{equation}\sigma \approx \gamma \frac{{\sum }_{i}^{{N}_{\mathrm{s}}}{n}_{i}}{S},\end{equation} \tag{ 2 }$$

where γ is a constant in units of energy-per-atom, N_s the number of undercoordinated atoms per unit cell, n_i the number of Au broken bonds and S the area of the unit cell. Ab initio calculations predict that the product σS gives energies per atom equal to 55.6, 81.3 and 120.3 kJ mol⁻¹ for (111), (100) and (110) Au surfaces, respectively [47]. Hence, the resulting AuNP equilibrium shape corresponds to a truncated octahedron exhibiting mostly (111) and (100) facets. The effect of including the interaction with a complex chemical environment is rationalised by considering a net overall decrease in Au surface energy and a deviation from the linear behaviour presented in equation (2) (with γ → 0) [24]. This combined effect leads to the stabilisation of multi-faceted NPs [25].

The results in figure 6 indicate that the adsorption of a single Asn helps reducing the energetic cost necessary for surface stabilisation by an amount between 28.7 and 45.8 kJ mol⁻¹ (see table 2). In the zwitterionic state (figure 6 left panel), Asn interacts preferentially with the (111) surface with a free energy gain of 5.6 kJ mol⁻¹ with respect to the (110) surface. This suggests that, upon Asn adsorption, the Au(111) facet is further stabilised with respect to the other facets, leading to rather isotropic AuNPs whose shape resembles the truncated octahedrons extensively discussed in the literature [48]. This prediction is consistent with HR-TEM and electron diffraction measurements shown in figures 4(e) and (f).

For the anionic state (figure 6 right panel), the main difference with the previous case is a general increase in Asn adsorption energy showing a reversal in the stabilisation sequence (see table 2). In this case the amino acid is more strongly bound to the (110) surface (5.5 kJ mol⁻¹ with respect to (111)). In the context of equation (2), these conditions correspond to an overall stabilisation of Au surfaces, with an additional decrease in surface energy differences between closed and open facets. As a result, open facets are more likely to appear, which agrees with scattering data showing higher population of (311) facets for basic than for acidic pH conditions (figures 4(g) and (h)). This could also qualitatively explain the HR-TEM micrographs at pH 9 showing relatively amorphous AuNPs (figures 2 and 4(b) in the SI). We emphasise here that the interaction with anionic Asn induces global stability and near degeneracy of Au surfaces leading to changes in morphology and size of the resulting NPs. Indeed, stabilisation of more open surfaces could also be the reason why we observe larger NPs at pH 9 (figure 4(d)). Stable open surfaces occupy larger surface areas that grow with the radius (square) of the NP.