# Highly efficient and selective extraction of gold by reduced graphene oxide

### Highly efficient gold extraction by rGO

The rGO was obtained by chemical reduction of commercial, mass-produced graphene oxide (GO) nanosheets using ascorbic acid as a reductant (Methods). The resulting rGO nanosheets were repeatedly washed with deionized water, until a stable colloid was formed, and then the rGO suspension was added directly into solutions containing gold ions (Fig. 1a). Considering the hydrochemical process is a flexible, low-cost, and sustainable method for e-waste recycling23,24, and [AuCl4] is a common gold complex seen in a hydrochemical process4,7,12, we have chosen KAuCl4 solutions prepared in different concentrations (C) to examine the gold extraction capacity of rGO (Fig. 1a, b); Similar to the behaviour observed for other adsorbents4,13,14, the extraction capacity increased with increasing C and reached a plateau above 10 ppm (Fig. 1b). Specifically, 1 g of rGO can extract ~ 690, 1180 and 1850 mg of gold from 0.1, 1 and 10 ppm gold solutions, respectively. This extraction capacity outperformed those of novel nanoporous adsorbents at the same concentrations and, importantly, extends to solutions containing only 1 ppb of gold (Supplementary Note 1; Supplementary Fig. 1 and Supplementary Table 1). Although we could not quantify the extraction capacity for even lower C (our quantitative methods reached their detection limit), we still observed significant gold extraction down to as little as 10 ppt (Supplementary Fig. 1). This makes the reported rGO technique particularly interesting because wastewater and freshwater often contain gold with C below 10 ppb, and its efficient reclaim is challenging12.

When measuring the extraction capacity of rGO to different pH from 2 to 11, we found the extraction capacity varied with the pH, which is not unusual and similar to many gold adsorbents4,13,14,15,16,19,20,21. As shown in Supplementary Fig. 2, the maximum gold uptake was found at pH ≈ 4. This is attributed to the fact that, at lower pH, the rGO colloid loses its stability25,26 whereas, at basic pH, rGO undergoes deprotonation, becomes negatively charged4 and starts repelling negatively charged [AuCl4]. Both effects decrease the extraction capacity. It is important to note that the extraction capacities at the acidic and basic solution are still higher than the previously reported gold extraction capacities at the same C (Supplementary Figs. 1 and 2), and by increasing the amount of rGO, ~99% gold ions from its 10 ppm solution with the pH of 2−11 can be extracted (Methods and Fig. 1b). We also investigated the gold extraction as a function of time (Fig. 1c and Supplementary Fig. 3). Within 10 min, 1 g of rGO provided extraction of 1010 and 325 mg of gold from 10 ppm and 100 ppb solutions, respectively (Fig. 1c), suggesting not only efficient but also rapid extraction. Interestingly, after extraction to 1 ppm gold, rGO changed its colour from black to gold within hours, suggesting almost instantaneous Au recovery (Supplementary Movie 1).

### Mechanism study of rGO’s gold extraction behavior

To understand the ultrahigh gold extraction behavior of rGO, we started with the characterization of rGO after the gold extraction. We found nanoparticles on top of rGO nanosheets after 24 h of its exposure to gold solution (Fig. 2a). X-ray diffraction (XRD) showed this particulate to be metallic gold, suggesting that [AuCl4] was reduced to Au0 (Supplementary Fig. 4). This was confirmed by X-ray photoelectron spectroscopy (XPS) analysis (Fig. 2b) which showed that >95% of the particulate was metallic Au0 rather than [AuCl4]. Thermal analysis (Supplementary Note 2; Supplementary Fig. 4) provided good agreement with the XPS results, as no endothermic peak for the transformation of gold salts into Au0 was found by differential scanning calorimetry (DSC). These observations suggest that a reductive adsorption mechanism dominates the extraction process and, if any adsorption of [AuCl4] on rGO occurs, its contribution is minor. Furthermore, we found that gold nanoparticles formed already after 2 min of the contact of rGO with gold solutions (Supplementary Note 3; Supplementary Fig. 7). Moreover, DSC analysis after 10 min showed no peak attributable to the gold salt precursor (Supplementary Fig. 8). This suggests that the reductive adsorption of gold ions happens throughout the entire extraction process, including its initial stages. In comparison to the reported novel gold adsorbents, though some showed reductive adsorption, other physi- and chemi-sorptions were also significant. That is, gold extracted by these adsorbents was a mixture of ionic and metallic gold4,12,13,14,15,19,27,28,29, therefore, it would require energy- and cost-intensive post-processing, for example, elution and precipitation to desorb and reduce the ionic gold, so that a full adsorption capacity can be achieved30,31. In contrast, as >95% extracted Au ions were reduced to metallic gold by rGO, it allowed direct isolation of metallic gold from rGO surface without further post-adsorption processing, providing an extra advantage of rGO for gold extraction.

Because no extra reductant for gold ions was used during gold extraction, it is sensible to conclude that the observed reductive adsorption was a redox reaction between the gold ions and rGO. We have performed gold extraction at different temperatures (T) to investigate the kinetics of this redox reaction. As the gold extraction capacity is a direct result of the redox reaction rate, we have plotted the extraction capacity versus T, and fitted with the Arrhenius equation, exp(–E/kBT), that E is the energy barrier and kB is the Boltzmann constant. This yielded an energy barrier 0.12 ± 0.02 eV (Fig. 2c), suggesting an activation behaviour for gold extraction, which is consistent with our DSC analysis that gold reduction by rGO is an endothermic process (Supplementary Fig. 4), the E is much lower than the activation energy seen in typical chemisorption process (0.21 eV-4.3 eV)32, supporting the observed fast and ultrahigh extraction capacity at low C. Furthermore, the observed temperature dependence allowed improving gold extraction capacity by increasing T, for example, at T = 45 °C and 60 °C, we observed extraction capacities (C = 10 ppm) of 4100 mg/g and 9059 mg/g respectively.

Next, we studied the influences of the redox pair, gold ion and rGO on the gold extraction performance. For the gold ions, we have measured rGO’s adsorption capacity to gold ions with different reduction potential E0 (Fig. 2d). Similar capacities were found for solutions of Au3+, [AuCl4] and [AuBr4]. Note that these complexes were most frequently used for gold extraction by novel adsorbents as well as for doping of graphene by gold-containing solutions4,12,14,21,33,34. This also suggested the possibility for rGO to extract gold from gold solutions containing different gold ions. As a demonstration, we found rGO reclaimed ~100% gold from 10 ppm [AuBr4] solution obtained by dissolving gold in the aqueous solution of N-bromosuccinimide at pH = ~835. For ions with E0 < 0.8 eV, for example, [Au(CN)2] and [Au(S2O3)2]3-, rGO showed little extraction, corroborating our conclusion that the redox reaction between gold ion and rGO dictated the reported behaviour.

To study the influence of rGO on the redox reaction, we recalled the local atomic structure of rGO nanosheet (Fig. 2h), which contains large unoxidized (or reduced) graphene areas, some oxidized regions that provide stability of rGO dispersions, and sub-nanometre pinholes36. Given a little fraction of pinhole (~5%) on rGO36, we only focused on the contribution of graphene area and oxidized regions to the gold extraction behaviour.

For the graphene area, in previous reports of gold ion (Au3+ or [AuCl4]) doped graphene grown by chemical vapour deposition (CVD), the graphene was found to chemically reduce these gold ions via a redox reaction mechanism, in which graphene donated electrons to reduce Au3+ to Au0.37,38 Though these studies only focused on a high C (≥200 ppm) and a short doping time (one to few minutes), they implied the observed reductive adsorption likely was accounted to the graphene area.

To validate that the graphene area of rGO reduced the gold ion, we have first studied the gold reduction by pristine graphene obtained by mechanical exfoliation, so that possible influence of the oxidized region can be excluded. The gold nanoparticles formed on the graphene after a few minutes and continued to grow even after many hours. Using areal extraction capacity allowed us to quantify the gold extraction capability of mechanically exfoliated graphene. The monolayer graphene showed the highest areal extraction capacity, in quantitative agreement with the weight capacity observed for rGO (Supplementary Note 4), suggesting the graphene area in rGO was responsible for ultrahigh gold extraction capacity. Rather unexpectedly, as shown in Fig. 2e, increasing the layer number of graphene (N), the areal gold extraction capacity, which should be independent of N as only the outmost graphene layer had contact with the gold ion, and there’s no gold intercalation between graphene layer, instead, decreased with N, indicating not only the redox potential between graphene and gold ions is essential, but also other factors are involved in the process. Further SEM analysis revealed that wrinkles and folds frequently seen for mono- and few-layer graphene tended to accumulate more gold nanoparticles than in flat areas (Supplementary Figs. 9 and 10), suggesting a higher chemical activity for gold extraction at the warped graphene, which was in agreement with previous reported gold doped CVD graphene37, this can be explained by our simulation (Supplementary Fig. 11) that the strain generated in the warped area of graphene, decreased the adsorption energy to gold ion and enhanced electron transfer of graphene to gold ion, both were beneficial for reductive adsorption of gold ion (Supplementary Note 4).

Further insights on the contribution of the graphene area to the gold reduction were obtained by considering the electron transfer from the graphene area to gold ions, the Raman and ultraviolet-visible (UV-Vis) spectroscopy were used to probe the electron transfer of rGO by comparing the spectra before and after gold extraction. The Raman spectrum showed that after extraction, the rGO became p-doped and showed an increased number of defects in the rGO nanosheets39 (Supplementary Figs. 5a and  6). The UV-Vis spectroscopy revealed that the characteristic peak of rGO shifted towards that of GO (Supplementary Fig. 5b). These two pieces of evidence indicated the donation of electrons from rGO for gold reduction40, confirming the reduction of gold ions by the graphene area of rGO. In another set of experiments, encouraged by the fact that graphene reductively adsorbed gold ions, the gold extraction capacities of commercial graphene and expanded graphite were measured. Both graphene materials were known to have well-retained graphene areas but little oxygen-containing functional group, which was also verified by the Raman analysis (Supplementary Note 5, Supplementary Fig. 12). During the extraction, they showed no dispersibility in gold solution because of their hydrophobicity, and their extraction capacities were <100 mg/g (Fig. 2g).

The low extraction capacity observed in commercial graphene suggested graphene area was not the only factor contributing to the ultrahigh extraction capacity of rGO. Thus, we studied the influence of the other atomic structure, oxidized regions of rGO, on the extraction performance. Given the synthesis of rGO was a process to remove some oxidized regions on GO (Fig. 2h), we have controlled the reduction time of GO, and XPS analysis confirmed the oxidized regions were gradually removed with the reduction time (Supplementary Note 5, Supplementary Fig. 12a). As shown in Fig. 2g, at a C = 10 ppm, pristine GO showed little gold extraction capacity, but the rGO obtained by 10-minute reduction exhibited an extraction capacity ~1100 mg/g, and 30-minute reduction gave the highest extraction capacity ~1850 mg/g (Fig. 2f), which could be explained by the increased graphene area in the rGO with a prolonged reduction time. Therefore, more sites were available for reductive adsorption. Further increasing the reduction time to 1 and 4 h, rGO agglomerated as validated by their decreased zeta potentials, and the capacities have decreased to 1460 and 1180 mg/g, respectively. This can be understood by that the reduction was beneficial for the recovery of graphene areas of rGO, but a long-time reduction excessively eliminated the oxidized regions. As a result, agglomeration and a poor dispersibility of the rGO were seen, such agglomeration of rGO disabled some of the recovered graphene areas from being exposed to the aqueous gold ion and inhibited the reductive adsorption, leading to a low extraction capacity (Fig. 2h). In addition, we also have tried rGO reduced by hydrazine or hydroquinone for gold extraction. Both rGO had recovered graphene area and formed stable monolayer dispersions as confirmed by their zeta potentials (Supplementary Note 5; Supplementary Fig. 12). They exhibited high extraction capacities (~1190 and 1835 mg/g from C = 10 ppm, respectively) as shown in Fig. 2g, similar to the values for the rGO reduced in ascorbic acid. These experiments unambiguously showed that both the graphene area and the oxidized regions were critical for the observed high gold extraction capacity.

With the above experimental results, we proposed the mechanism for the ultrahigh gold extraction of rGO as follows, firstly, the adsorption was driven by the concentration difference (ΔC) of adsorbates in its solution and on the adsorbent41. In our case, ΔC=$${C}_{{{{{\mathrm{sol}}}}}}^{{{{{\mathrm{Au}}}}\;{{{\mathrm{ion}}}}}}$$$${C}_{{{{{\mathrm{rGO}}}}}}^{{{{{\mathrm{Au}}}}\;{{{\mathrm{ion}}}}}}$$, Once gold ions were adsorbed on rGO, >95% of them were reduced to Au0 spontaneously. Such rapid and complete conversion of gold ion to Au0, left an extremely low $${C}_{{{{{\mathrm{rGO}}}}}}^{{{{{\mathrm{Au}}}}\;{{{\mathrm{ion}}}}}}$$, consequently, a high ΔC was maintained during the adsorption process. Therefore, rGO overcame the fast equilibrium at low C, and showed an ultrahigh adsorption capacity at even ppb level. Secondly, the reductive adsorption was enabled by the graphene area of rGO, it donated electrons to the adsorbed gold ion, which was reduced to Au0. In addition, the wrinkles and warped area of graphene prompted the adsorption of gold ions and electron transfer to gold ions during the redox reaction, further increasing the gold extraction capacity. The oxidized regions of rGO, provided a good dispersibility of rGO, therefore allowing efficient adsorption and reduction of gold ions by graphene area, leading to an ultrahigh extraction capacity (Fig. 2h).

### Realization of high selectivity for gold extraction

Based on the described understanding, we explore the possibility of extracting Au from e-waste that typically contains a variety of other metals. Initial tests were performed using a simulated e-waste mixture containing [AuCl4], Cu2+, Ni2+ and [PtCl4]2- ions. We observed ~99% gold recovery with <5 % of the other metals being extracted (Supplementary Fig. 13). Next, a discarded central processing unit (CPU) was leached with aqua regia, forming leachate containing ~88 ppb of gold ions. Our rGO colloids captured ~90% of this gold (Fig. 3b) but also took up co-existing ions, similar to the other previously used gold adsorbents4,10,12,14. The uptake was from 7 to 100 % for 6 out of the 14 metals present in the CPU leachate in concentrations from sub-ppm to hundreds of ppm (Fig. 3b).