Characterizing Self-Assembled Monolayers on Gold Nanoparticles

A key aspect of nanoscience is to control the assembly of complex materials from a “bottom-up” approach. The self-assembly and self-organization of small ligands at the surface of nanoparticles represent a possible starting route for the preparation of (bio)nanomaterials with precise (bio)physical and (bio)chemical properties. However, surface characterization and elucidation of the structure-properties relationship, essentials to envision such control, remain challenging and are often poorly investigated. This Topical Review aims to discuss different levels of surface characterization, giving an overview of the experimental and computational approaches that are used to provide insights into the self-assembled monolayer with molecular details. The methods and strategies discussed focus on the characterization of self-assembled monolayers at the gold nanoparticle surface, but most of them could also be applied to other types of nanoparticles.


Introduction
The Bakerian Lecture given by Michael Faraday in 1857, describing for the first time the interaction of metal particles with light, has been a milestone in the development of the modern science of gold colloids. 1,2 The enticement of gold nanoparticles (GNPs) resides in the way their free electrons interact with light, 3 revealing both their and the environment's properties, together with the ease of synthesis 4,5 and functionalization with small ligands. 6 Thus, small thiolate ligands, including alkanethiols, 7 peptides, 8 oligonucleotides, 9,10 and polyethylene glycols, 11 bind readily to the surface and form self-assembled monolayers (SAMs), 7 imparting both stability 12,13 and desired functionality 14,15 to the nanoparticles. However, even though the literature is abounding with examples of functionalized gold nanoparticles for applications in biology, 16 catalysis 17 and sensing, 18 usually illustrated with inspiring schemes, the actual structure and organization of the SAMs at the gold nanoparticle surface are challenging to assess and remain often poorly characterized. Such elucidation is a fundamental prerequisite to warrant data reproducibility and build knowledge of predictable properties on the basis of the nanomaterials' characteristics. Moreover, the investigation of the relationship between structure and function is necessary to fully exploit the possibility of engineering nano-objects with well-defined (bio)physicochemical and structural properties from a "bottom-up" approach, which could then be assembled into complex networks, potentially taking inspiration from nature. 19,20 Starting from basic properties, such as overall chemical composition and thickness, and progressing to more advanced insights, such as molecular structure and phase separation of ligands, we review the current knowledge and the methods available to characterize these systems.

2
As the Topical Review is organized according to self-assembled monolayer properties, the same studies may appear in different sections. Overall, this Topical Review aims to highlight the web of connections existing between the different self-assembled monolayer characteristics.

Basic Characterization and Composition
As mentioned in the introduction and also highlighted in previous reviews, 7,21 many scientific articles lack basic characterization of functionalized GNPs. Thus, in this section, some of the chemical and analytical approaches and techniques adopted to characterize a SAM on GNPs and shed light on its composition are recapitulated.
Once the GNPs are synthesized, the stabilizing molecules at the surface, e.g., citrate ions 4 or simple thiols, 5 can be replaced by another thiol-containing molecule in a ligand exchange reaction, following an associative reaction with 1:1 stoichiometry. 22 The stabilizing molecules ensure colloidal dispersion of the GNPs, typically by electrostatic and/or steric repulsion. Scheme 1 lists the most common experimental methods used to characterize the ligands bound to the nanoparticle surface, which will be discussed throughout this Topical Review. Such methods can be divided into those that analyze directly the ligand-capped GNPs and those that require the release of the ligands (Scheme 1).
The first category includes UV-visible spectrophotometry. Indeed, GNPs larger than 2 nm in diameter are characterized by the surface plasmon band (SPB), 23 theoretically described by Mie in his work dating from 1908; 24 the position of the SPB maximum is deeply influenced by the GNP size and shape and the dielectric constant of the surrounding environment. Hence, the presence of ligands anchored onto the surface can be indirectly probed by a shift in the SPB maximum detected simply by UV-vis spectrophotometry. Scheme 1. Examples of experimental methods used to characterize the GNP's surface, analyzing ligand-capped GNPs (left) or ligands released upon gold decomposition or ligand exchange (right).
Also, indirect evidence of functionalization can be obtained from the increased colloidal stability in high ionic strength media that otherwise would lead to complete particle-to-particle aggregation, i.e., red-shift and broadening of the SPB. 25 Weisbecker et al., for instance, investigated the effect of capping citrate GNPs with alkanethiols of different chain lengths and terminal functional groups in various conditions of pH and ionic strength by monitoring the absorbance spectra over time; 26 Lévy et al. observed a 2 nm red-shift and protection to salt-induced aggregation 3 resulting from the conjugation of the CALNN pentapeptide to citrate GNPs ( Figure 1A). 8 Duchesne et al. observed different GNP stability against electrolyte-induced aggregation and ligand exchange upon varying the ligand concentration during the conjugation process, thus providing an indirect measurement of the ligand capping density on GNPs. 12 Evaluation of the resistance of capped GNPs to increasing concentrations of dithiotreitol, i.e., ligand exchange, was used to select combinations of ligands forming highly compact SAMs on GNPs, hence providing high stability in biological environments. 13 Similarly, indirect estimation of the SAMs' compactness was gathered by Schulz et al. from assessing the resistance to cyanide etching of GNPs capped with SAMs constituted of PEG-thiol ligands with the same chain length, but with the segments between the thiol group and the polymer moiety of different sizes and chemical functionalities ( Figure 1B). 27 Furthermore, differential centrifugal sedimentation 28 (DCS) or dynamic light scattering (DLS) and zeta (ζ) potential measurements 29 are used to determine the hydrodynamic diameter and the charge of the ligand-capped GNPs, respectively, thus giving insight into their stability and properties, such as protein 30 or molecule 31 adsorption. Thermal gravimetric analysis (TGA), by monitoring changes in weight over a temperature ramp, gives an indication of the amount of ligands on the GNPs, but requires 1-10 mg of sample and assumptions and modelling for the shape of the gold cores. 32 IR and NMR spectroscopies are generally used not only to characterize the content of SAMs on GNPs, but also to investigate their structure and interactions existing between the ligands; thus, they will be further discussed in Section 4.
Examples of analytical methods to identify and quantify the amount of ligands upon cleavage from the GNPs surface include Zhou et al.'s work where SAMs constituted of up to three different ligands were first cleaved from the NP surface with I 2 , 33 then identified by mass spectrometry and quantified by high-performance liquid chromatography with UV or chemiluminescent nitrogen detection (HPLC/MS/UV/CLND), using calibration curves constructed from the free ligands. 34 This provided a method to confirm the common assumption that the ratio of ligands on the GNP surface is the same as the one used during the conjugation process ( Figure 1C). However, as Fisher et al. pointed out, this method is limited to disulfide ligands and cannot quantitatively analyze mixtures of simple alkanethiol ligands because of disulfide formation upon I 2 exposure. 35 Thus, Fisher et al. developed a method to quantify the amount of alkanethiol ligands in mixed SAMs on 2 nm GNPs using gas chromatography coupled to mass spectrometry (GC/MS/MS), upon I 2 exposure, and using a calibration curve constructed from free disulphide mixtures; 1 H-NMR spectroscopy was employed as an additional technique to corroborate their findings. 35 Hinterwirth et al. 36 described a method to determine the ligand capping density on the basis of the observed linear relationship between the NP diameter, measured by transmission electron microscopy (TEM), and the ratio between gold and sulfur atoms, measured by inductively coupled plasma mass spectrometry (ICP-MS), a powerful technique for elemental analysis. 37 The number of peptide ligands can also be measured using amino acid analysis. 8,12 To quantify the amount of alkanethiol-oligonucleotides capping a GNP, Demers et al. modified the ligand by fluorescently labelling it, and then, after conjugation to the GNPs and removal of the ligand excess, incubated the capped GNPs in the presence of another thiolate ligand. Thus, the fluorescently labelled oligonucleotides were released by ligand exchange and quantified using fluorescence, based on calibration curves made from the free fluorescently-labelled ligand in solutions of same pH and ionic strength ( Figure 1E). 38 The ligand exchange reaction was   where the GNPs were capped with a mixed SAM constituted of a ω-carboxylate and ω-amino thiols; the last one functionalizable with a fluorescently labelled NHS ester derivative ( Figure 1F). 40 Functionalized GNPs can also be characterized by laser desorption ionization mass spectrometry (LDI-MS). [41][42][43] The use of this technique for characterizing nanoparticles originates from Tanaka et al.'s work, 44 demonstrating the ability of metal nanoparticles to absorb laser light energy and thus desorb analytes and enhance their ionization.
For instance, Yan et al. characterized GNPs capped with positively or negatively charged and neutral ligands by LDI-MS: in all cases, both thiol and disulphide ions were detected. 39 Moreover, when analyzing GNPs functionalized with two different ligands, not only thiol and disulphide ions corresponding to both ligands were identified, but also mixed disulphide ions ( Figure 1D). The authors developed a procedure to provide a semi-quantitative measurement of two ligands on GNPs; since different ionization efficiencies are associated to different ligands, NMR spectroscopy was used as external technique for calibration.
Cliffel and McLean's group developed a strategy coupling MALDI to ion mobility-mass spectrometry (IM-MS), a gas-phase separation technique, to provide a measurement of the relative quantities of two ligands in a mixed SAM on GNPs. 45 This technique is more advantageous than NMR spectroscopy, since it does not require the two ligands to have different functionalities. Also, unlike traditional MS, IM is able to separate ions corresponding to clusters of gold atoms and ligands from those corresponding to ligands alone, thus proving that the ligands were actually attached to the GNP surface.

Compactness and Thickness
As discussed in the previous section, different approaches have been proposed in the literature to determine the amount of ligands on GNPs, hence the capping density. Whilst not independent of the capping density, the SAM compactness and thickness also depends on the conformation of the ligands and on the amount of solvent present within the SAM. In this section, direct and indirect methods providing insights into these characteristics are detailed.
The compactness of a SAM does not only result from the number of ligands on the GNPs surface; Chen et al.
have indeed reported the extremely high stability against ligand exchange of SAMs constituted of peptidols with aromatic amino acids, i.e., H-CFFFY-ol or H-CFFFT-ol, and PEGylated alkanethiol, indicating that aromatic and alkyl chains can form extremely compact SAMs through hydrophobic interactions and, possibly, well-defined secondary structure motifs. 13 Duchesne et al. highlighted that higher SAM compactness can be achieved by using modified peptides, i.e., peptidols, which bear a non-ionisable primary alcohol group, instead of a carboxyl group, at the C-terminus. 12 Schulz et al., as previously mentioned, probed the effect which segments of different sizes and chemical functionalities between the thiol group and a polymer chain had on protecting the GNPs from cyanide etching, thus showing the effect of the segments on the capping density, hence compactness and stability to etching ( Figure   1B). 27 The packing of a SAM can be measured more directly via its thickness, which, however, can be challenging to assess. Since GNPs are electron dense materials, they have been used for over 60 years as contrast agents in transmission electron microscopy (TEM). 46 Even though TEM is mostly used to size the gold cores only, Jürgens et al.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 6 measured the thickness of films of proteins adsorbed on GNPs with this technique. 47 Ardao et al. imaged by TEM samples of enzyme-adsorbed GNPs, not purified from the excess of enzyme, using the negative staining method. 48 The agent, by staining the enzymes' surface, conferred a grey background to the images. Since the enzymes at the GNPs' surface were highly packed and thus did not get stained, the ligand shells appeared as white rings in the TEM images. However, considering the drying of the sample and the often insufficient contrast, TEM images cannot provide quantitative information on the thickness of SAMs in solution.
As mentioned in Section 2, dynamic light scattering (DLS) is a common method to determine the hydrodynamic diameter of GNPs, 29 but it requires rigorous experimental conditions in order to avoid data misinterpretation. 49 Its value depends on the length of the ligands and the compactness of the monolayer. Therefore, changes of these parameters can be reflected by DLS measurements. However, small changes in the hydrodynamic diameter (<5%) cannot be detected by DLS. Walkey et al., for instance, assessed by DLS the hydrodynamic volume occupied by one poly(ethylene glycol) (PEG) molecule in SAMs of different capping densities on GNPs of different sizes. 50 Smaller volumes were found for higher capping densities, indicating both strengthening and dehydration of the ligands. On the other hand, for a given capping density, a larger volume was found for smaller GNPs, which was rationalized with the fact that a decrease in GNP size, i.e., an increase of curvature, frustrates the hydrophobic interactions between the PEG chains ( Figure 2A). Belsey et al. described two approaches to measure the thickness and derive the number of either adsorbed proteins or conjugated peptides on GNPs of different sizes. 51 The first approach involved UV-Vis spectroscopy to indirectly prove the adsorption/conjugation of the biomolecules to the GNPs, by monitoring the surface plasmon band, DLS to measure the thickness of the monolayers, and differential centrifugal sedimentation (DCS) to prove the monodispersity of the GNPs. The second approach employed X-ray photoelectron spectroscopy (XPS) to measure the monolayer thickness in ultrahigh vacuum. The authors concluded that in spite of the intrinsic differences between the two approaches, the results appeared to be consistent. Also, Torelli et al. developed a computational model for determining the ligands' capping density from XPS data for poly(ethylene glycol) alkanethiol-capped GNPs of different sizes. 52 Falabella et al. 53 and Krpetić et al. 28 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58

Conformation
As discussed in the previous section, measurement of the SAMs' thickness and compactness also provides insights into the underlying ligands' conformation. However, determining the structure of ligand-capped GNPs at the molecular level remains highly challenging.
The work by Jadzinsky et al., dating from 2007, marked a breakthrough in terms of structural information on GNPs. 54 The authors determined the crystal structure at 1.1 Å resolution of p-mercaptobenzoic acid-capped 1.5 nm gold clusters and thus, shed light on the structure and geometry of a thiol monolayer. Azubel et al., in their work dating from 2014, solved the structure at atomic resolution of a smaller thiol-capped gold cluster by aberrationcorrected transmission electron microscopy. 55 In the years in between these two works, capped GNPs in the 1-3 nm range were crystallized, but X-ray diffraction structures with high resolution were not obtained. Therefore, even though the structures of gold clusters that have been effectively solved provide precious information, other approaches have to be sought to elucidate the conformation of ligands on GNPs larger than 2 nm.
Examples of such attempts are described in this section.  56 Using a combination of UV-vis and DLS techniques, the authors found that the changes in GNP properties were correlated with different PEG conformations. The hydrodynamic radius of the functionalized GNPs was found to suddenly increase at a ratio between PEG and MUA of ~0.7, marking the straightening of the PEG ligands into a brush-like conformation. The blue-shift of the surface plasmon band (SPB) maximum at this ratio corroborated the change in PEG conformation. On this note, Tagliazucchi et al. presented a theoretical method to calculate the optical properties of GNPs capped with SAMs of different conformations, which have different densities and hence refractive indices, and thus were able to correlate the position of the SPB with a change in conformation. 57 The theoretical results were found to well describe the optical properties of GNPs functionalized with a thermoresponsive polymeric monolayer ( Figure 3A), i.e., the red-shift of the SPB upon polymer collapse, as well as the effect of capping density, chain length and GNP size on the extent of the shift. Xia et al. characterized by DLS gold nanocages with side lengths of 30, 50 and 60 nm, functionalized with PEGs of different molecular weights. 58 The largest hydrodynamic diameter corresponded to the GNPs capped with the PEG molecules of intermediate size.
Murray and co-workers characterized by FTIR GNPs from 1.5 to 5.2 nm in diameter capped with dodecanethiolate ligand. 32 The IR spectra were taken in solid-phase and allowed the investigation of the alkyl chain conformation. More ordered SAMs, corresponding to a trans zigzag structure, similar to the structure observed on planar gold surfaces, were found on the larger GNPs. However, considering the relatively small size of the largest GNPs and that ordered monolayers were already distinguished on 4.4 nm GNPs, the authors investigated the role of sample preparation in such observations. Therefore, the structure of alkanethiols of different chain lengths on ~2 nm GNPs was analyzed by FTIR both in solid-phase and in solution. 33 It was found that shorter alkanethiols adopted a highly ordered conformation on these 2 nm GNPs in solid-phase, but became quite disordered in carbon tetrachloride, highlighting the importance of the sample preparation, as it can lead to changes in the ligands conformation. Moreover, Murray and co-workers listed the factors determining the broadening of thiolate-capped GNP 13 C and 1 H NMR spectra. 32,60 Importantly, since both the ligands' proximity to the gold surface and the density of packing contribute to spectral broadening, NMR analysis provides information about the ligands' environment and structural features. The spectra, however, can be difficult to interpret.
Peptide-capped GNPs have been described by us as building-blocks that could potentially be engineered into advanced nanoparticles with protein-like properties. 19 Hence, determining their structures is of crucial importance.
Fabris et al., prepared GNPs of ~1-2 nm size, capped with peptides made of α-aminoisobutyric acid residues. 61 Homooligomers of this residue (starting from the tripeptide) are known to form stiff secondary structure motifs, i.e., β-turns and 3 10 -helices. IR analysis indicated the presence of 3 10 -helices, stabilized by intra-peptides hydrogen bonds, both in solution and at the surface of GNPs, independently of the peptide length. Also, the authors judged inter-peptide hydrogen bonds, identified by NMR analysis, to underlie the rigid conformation observed in all the systems investigated. Thus, building-blocks characterized by a stiff monolayer and a precise secondary structure motif were provided. It is worth noting that the IR spectra of the capped GNPs, taken both in solid-phase and in solution, did not show any significant difference.
Shaw et al. investigated the effect of GNP size, i.e., curvature, on the secondary structure of CALNN and CFGAILSS peptide SAMs. 62 The authors used GNPs of 5, 10 and 25 nm in diameter, and a combination of FTIR and solid-state NMR (ssNMR) to elucidate these structures. The CALNN peptide was found to adopt a random coil conformation independent of the GNP size. The CFGAILSS peptide, which arranged in amyloid fibers with an antiparallel β-sheet conformation in solution, formed parallel β-sheets when attached to the gold surface, which were more pronounced on the largest GNPs. Hence, the number of adjacent peptides forming inter-peptide hydrogen bonds and thus, parallel β-sheets, is dictated by the nanoparticle curvature, which agrees well with a simple geomet-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 10 content of α-helix motif on the GNPs with a lower curvature, similar to the one observed for the free peptide or on a planar gold surface, whereas GNPs with high curvature were dominated by β-sheet structures ( Figure 3B). The authors suggested that this was due to high capping densities at the edges and corners of GNPs, caused by the high unsaturation of the gold atoms. High capping densities, in turn, allow for the efficient formation of inter-peptide hydrogen bonds which are the prerequisite for forming β-sheet structures. Since in small GNPs, edges and corners dominate over planar faces, more β-sheet structures were found for these, whereas in 20 nm GNPs, binding to planar faces seemed to be predominant, leading to the native α-helical peptide structure. Thus, both works highlighted the importance of GNPs size and ligand sequence when designing nanomaterials with well-defined secondary structure motifs.
On a different note, Rio-Echevarria et al. confirmed by circular dichroism (CD) that the helical secondary structure of an undecapeptide was maintained after conjugation to GNPs of 1-2 nm. 63 However, measuring the CD of larger GNPs in the UV region (to elucidate the conformation of the peptide SAM) can be problematic due to the high absorbance of the gold cores.

Organization of Mixed Self-Assembled Monolayers
Since the ratio of ligands at the GNPs surface does not necessarily reflect the one used during conjugation, in Section 2 we summarized some of the efforts made to quantify the ligands constituting a mixed SAM on GNPs.
However, questions related to the organization of mixed SAMs on GNPs should also be addressed.
In the 90s, studies of wetting, performed with ellipsometric and goniometric measurements, suggested that two alkanethiols of different chain length and with either a methyl or hydroxymethyl terminal functional group phaseseparate on a planar gold surface, forming microscopic islands ( Figure 4A). 64 Topological characterization with scanning tunneling microscopy (STM) by Stranick et al. 65 confirmed such phase separation into domains, whose shapes were found to be time-dependent, for two alkanethiols of the same length, but having different noninteracting terminal groups, i.e., methyl and methyl ester, on a gold surface. Tamada et al. 66 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 12 analyses received a response from the authors of the original publication that addressed the arguments of invalidation of the results obtained. 73,74 Other approaches to assess the organization of mixed SAMs include Duchesne et al.'s work where GNPs were capped with CALNN and a functional peptide, containing two reactive groups, and acting as a proximity probe. 67 Intermolecular cross-linking between two functional peptides was expected if their separation was lower than 4 nm; if greater, intramolecular cross-linking should have been detected ( Figure 4B). Using this method, the authors showed that even if there were only two or three functional peptides on a GNP, they had the tendency to be close to each other on the GNP surface; i.e., they organized in a small domain. This packing was found to form during the self-assembly of the mixed monolayer, and not when the functional peptides were still free in solution. Prins' group used a range of probes to assess the organization of mixed SAMs. 75,76 In a first approach, the SAMs were constituted of different ratios of thiolate ligands with either a cationic or neutral terminal group, where the positively charged group was able to catalyse the transphosphorylation of a substrate, and the apparent catalytic rate constant k cat was used to probe the SAMs' organization. 75 The GNPs' catalytic activity was found to gradually increase up to ligand ratios of ~0.4 and then to remain constant; this increase was ascribed to a random distribution of the ligands within the SAMs. In ref. 76 , the same authors discussed the disavantages of k cat as parameter and proposed a general methodology to assess the morphology of a mixed SAM, without the need of a catalytic ligand. A fluorescent anionic probe was used, whose fluorescence is quenched upon interaction with the GNP surface, thus allowing for easy determination of its binding to the cationic ligands, the extent of which depends on the ligands' organization on the GNP surface. A random distribution of the ligands up to ratios of ~0.4 was concluded from both methodologies and was supported by simulated saturation profiles. Gentilini et al. observed a different surface organization and ordering in mixed SAM on GNPs they investigated. They used a radical probe to assess the organization of mixed SAMs constituted of perfluoroalkyl and alkyl amphiphilic thiolate ligands ( Figure 4C) on GNPs by electron spin resonance (ESR) spectroscopy. 68 The radical probe was able to sense the hydrophobicity of the environment and was characterized by different hyperfine splitting constants when sensing a SAM constituted of either perfluoroalkyl or alkyl thiolate ligands. Thus, the authors showed that, up to a ratio of 2.5 between alkyl and perfluoroalkyl ligands, the probe sensed a fluorinated environment, indicating the packing within islands of such ligands.
Cliffel and McLean's group used MALDI coupled to ion mobility-mass spectrometry (IM-MS) to investigate the phase segregation of two ligands in a mixed SAM on GNPs. 69 Upon MALDI analysis, gold-thiolate ions were generated and, once separated from the ligands' ions by IM-MS, the information regarding a particular type of goldthiolate clusters, i.e., Au 4 L 4 , was extracted and quantitatively compared to a theoretical model based on the binomial distribution. Using this method, the authors, characterizing several mixed SAMs of different compositions and ratios, observed various degrees of ligands segregation ( Figure 4D).
Additionally, from the analysis of the cross peaks in two dimensional NMR nuclear Overhauser effect spectroscopy (NOESY) spectra, information on the nucleis' proximity can be obtained. Thus  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 13 ously mixed SAMs. 78 Similarly, Liu et al. used a combination of one and two dimensional NMR to elucidate the organization of aliphatic and aromatic thiols at different ratios on GNPs. 79 Finally, efforts made to impart different chemical functionalities or polarity to the two hemispheres of an organic or inorganic core are worth mentioning. Such category of micro-and nanoparticles was described for the first time with the term "Janus" by De Gennes in 1991. 80 We refer the reader to past reviews highlighting the various procedures investigated to insert elements of dissymmetry within the monolayer. 81,82 The most common method involves an anisotropic environment, e.g., the interface between air/water or immiscible liquids, that drives the selforganization of a SAM constituted of hydrophilic and hydrophobic ligands. 83,84

Modelling
We refer the reader to a recent review recapitulating the different computational methods available to gather a better description and understanding of the interactions between a gold surface or nanoparticle and the surrounding biological environment. 85 In particular, here we consider the limited insights into the SAM's structure and organization that can be obtained with experimental techniques, and the information about SAMs on GNPs that can be obtained with molecular details using a combination of experimental and computational approaches.
Considering that alkanethiol-capped GNPs have been investigated since the early '90s, it is not surprising that several classical force fields (FF), describing the interactions of alkanethiols on gold, are reported in the literature. [86][87][88][89] Glotzer and co-workers reported for the first time an atomistic molecular dynamics (MD) study investigating the structure of alkanethiols on gold nanoparticles as a function of temperature, length of the ligand and GNP size; the sulfur atoms were fixed at a distance of 0.238 nm from the gold surface. 90 Ligands with 13 carbons on a 7 nm GNP did not exhibit long-range order at high temperatures, i.e., 900 and 600 K, while they did group into bundles with the same angle to the surface normal at low temperatures, i.e., 450 and 300 K. At 300 K, long-range order was reported only for alkane thiols with more than nine carbons. At high temperatures, the monolayer constituted of the ligands with 13 carbons was found to be more ordered on the larger GNPs, as illustrated by a decrease in the smallest average distance between adjacent terminal groups with an increase in GNP size.
Moreover, Glotzer and co-workers 91 performed atomistic molecular dynamics and mesoscale dissipative particle dynamics simulations to investigate the origin of the stripe-like patterns observed by Jackson et al. 70 They modelled the GNP surface as a sphere and simulated the GNPs capped with an equimolar mixture of immiscible alkanethiols, varying the length and the functional terminal group of the ligands, and the GNP size. According to the authors, a sufficient gain in conformational entropy was needed for the formation and stabilization of the stripe-like patterns. The entropy gain, due to the additional ligand-ligand interfaces in comparison to a bulk-separated system like the Janus one, was found to be dependent on the difference in length between the ligands, on the terminal functional groups, and on the GNP curvature. Furthermore, the authors argued that in case of sufficient entropy gain, the stripe-like patterns were corresponding to the morphology at the equilibrium. For instance, while stripe-like patterns were found on GNPs capped with a mixture of ligands with well-defined length ratios or with different terminal functional groups, a Janus organization was observed on small GNPs or in SAMs constituted of two short alkanethi-  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 14 ols with small terminal functional groups and similar length. In a subsequent work, Glotzer and co-workers used atomistic molecular dynamics simulations to study the factors influencing the thickness of the stripe-like motifs. 92 Velachi et al. 93 and Van Lehn et al. 94 performed atomistic MD simulations at room temperature and in explicit solvent to investigate the relationship between the arrangement of alkanethiols and the hydration properties of mixed SAMs constituted of equimolar alkanethiols of different lengths with either a hydrophilic or hydrophobic terminal functional group on GNPs of different sizes. The sulfur atoms were fixed on the gold surface and three pre-defined types of surface morphology were considered, namely, random, stripe-like (or patchy) and Janus arrangements.

Conclusions and Perspectives
The ability to control the self-assembly and self-organization of small molecules at the surface of nanoparticles could pave the way to the preparation of nanomaterials with well-defined structural and (bio)physicochemical properties, which could then be envisioned to form more complex systems. Whilst getting a detailed picture remains challenging, a number of characterization techniques are available.
For instance, the optical and electronic properties of the gold core can be exploited to gather indirect evidence of ligand conjugation with UV-vis spectroscopy, whereas, upon ligand removal, separation and mass spectrometry techniques can provide more direct information about the monolayer composition. Scattering techniques can give information about the whole self-assembled monolayer, thus tackling the questions of the nanoparticles' hydrodynamic radius, stability, size dispersion and monolayer thickness. The ligands' conformation is at the basis of these observations and attempts at assessing it involve structural biology techniques, such as NMR and IR spectroscopies; sample preparation procedures have to be carefully taken into account when interpreting the results, though.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 16 However, the level of molecular details that can be obtained with direct methods of characterization is limited.
Thus, using a combination of experimental and theoretical approaches can provide a compelling description of the structure and organization of the monolayer. Nevertheless, advances in both areas of research are critical to work toward the rational design of nanoparticles with precise structural and (bio)physicochemical properties.