An exhaustive description relative to the modeling of the different BNNTs used in this work in their pristine state is available in a very recent work by us [18]. The BNNTs are automatically generated by the CRYSTAL code through the wrapping of a periodic boron nitride monolayer (hexagonal P3m1 layer symmetry group) into cylinders of different radius and fully exploiting the symmetry operators of the nanotubes [36,37,38,39,40]. In the interest of the present work, it is worth mentioning that the calculated electrostatic potential maps indicate a prominent positive/negative valued potential region for the (4,0) BNNT, which become progressively less pronounced for BNNT with increasing radius, until obtaining a practically shallow electrostatic potential for the (15,0) BNNT. For the interaction with Gly, the geometry optimizations were carried out as 1D polymers within the P1 space group, in which the unit cell parameters have been enlarged twice to avoid lateral interactions between molecules of adjacent unit cells.

The different optimized adducts for the adsorption of Gly on the BNNTs are shown in Figure 1, whereas the calculated adsorption energies alongside the pure electronic and dispersion contributions are shown in Table 1. The adsorption energies (∆E_ads) per unit cell of the probe molecules with the BNNTs are computed as: ∆E_ads = E(Gly/BNNT) − E(BNNT) − E_m(Gly) (1) where E(Gly/BNNT) is the energy of a fully relaxed unitary cell containing the BNNT in interaction with Gly, E(BNNT) is the energy of a fully relaxed unitary cell of the BNNTs alone, and E_m(Gly) is the molecular energy of the free Gly.

For all the BNNT systems, six initial structural guesses were considered (see Figure 1a): (i) pure interaction between the NH₂ group of Gly and one nanotube B atom (hereafter referred to as BN/NH₂); (ii) BN/NH₂ interaction plus H-bonding between the Gly OH group and one nanotube N atoms (hereafter referred as BN/NH₂–OH); (iii) and (iv) interaction between the Gly CO group and one nanotube B atom plus H-bonding between the Gly OH group and one nanotube N atom, with the difference that the nanotube B and N atoms are chemically bonded or not (hereafter referred to as BN/COOH-1 and BN/COOH-2, respectively); (v) interaction of Gly in its zwitterionic form (hereafter referred to as BN/zwitt), in which the Gly COO⁻ group interacts with one B atom and the Gly NH₃⁺ is H-bonded to one nanotube N atom; and (vi) interaction between Gly and the BNNTs purely through the π system of the Gly COOH group (hereafter referred to as BN/π).

Data reported in Table 1 clearly indicate that the most stable adduct for the (4,0) BNNT is the BN/NH₂–OH whereas for the (6,0) BNNTs both BN/NH₂ and BN/NH₂-OH are nearly degenerate. These complexes result from dative covalent interactions between the Gly NH₂ group and the B atoms, which act as Lewis acid sites. These findings are consistent with the data reported for the interaction of these BNNTs with probe molecules, in which the interaction of NH₃ with the (4,0) and (6,0) BNNTs was found to be the strongest one among all tested molecules [18]. The fact that the BN/NH₂-OH adduct becomes the most stable one for the Gly/BNNT(4,0) system is consistent with the large polar character of this nanotube, the N atoms acting as H-bonding acceptor groups. This is not in line with the most stable BN/NH₂ adduct found for the (6,0) BNNT and is due to the weaker H-bonding acceptor character of the N nanotube atom when increasing the nanotube radius, which is reflected by an increase of the H-bond distance in the BN/NH₂–OH adducts (1.686 and 1.748 Å for the (4,0) and (6,0) BNNTs, see Figure 1a). In the same way, the B-N_Gly bond lengths of the dative interactions in the BN/NH₂ and BN/NH₂–OH complexes also increase with increases in the nanotube radius due to the progressive decrease of the Lewis acceptor character of the B atom. Because of that, the calculated adsorption energy (∆E_ads) is more negative and larger for the most stable BN/NH₂–OH adduct of Gly/BNNT(4,0) than the most stable BN/NH₂ adduct of Gly/BNNT(6,0) (i.e., −33.2 and −18.9 kcal mol⁻¹, respectively). The calculated energetic contributions; i.e., purely electronic and dispersion (∆E_el and ∆E_D values of Table 1, respectively) indicate that Gly adsorption on the (4,0) BNNT is largely dictated by the covalent dative interaction, whereas on the (6,0) BNNT ∆E_el decreases in favor of dispersion. An analysis of the Mulliken charges (Q values of Table 2, only limited to the most stable Gly/BNNT complexes) confirms the formation of charge transfer complexes for both the Gly/BNNT(4,0) and Gly/BNNT(6,0) systems, the computed charge transfer values from Gly to the BNNTs being 0.30 e and 0.22 e, respectively.

The most stable Gly/BNNT(15,0) adduct has been found to be the BN/π one. This is in perfect agreement with the EPM results, which point out the (15,0) BNNT as a practically non-polar nanomaterial. Data reported in Table 1 clearly indicate that the binding mechanism involved in this adduct is mainly based on dispersive forces (calculated ∆E_ads is practically equal to ∆E_D; i.e., −10.2 and −11.7 kcal mol⁻¹, respectively) dictated by π-stacking interactions between the π systems of the COOH group and the B-N hexagon rings of the (15,0) BNNT. Because of the presence of only non-covalent interactions, ∆E_ads is less negative compared to Gly interaction on (4,0) and (6,0) BNNTs. It is worth mentioning that the very same BN/π complexes have also been calculated for the Gly/BNNT (4,0) and (6,0) systems, meaning that, in the former case, the structure collapses onto the BN/NH₂ complex, whereas for the latter case the calculated ∆E_ads is found to be 9.0 kcal mol⁻¹ above the most stable one, due to the lower propensity of this BNNT to establish π-stacking interactions.

The interaction of Gly with the (9,0) BNNT is a frontier case between small radius (i.e., (4,0) and (6,0)) and large radius (i.e., (15,0)) BNNTs. Although the BN/NH₂ adduct has been found to be the most stable one, the BN/π complex is the second most stable one lying 1.9 kcal mol⁻¹ above. Calculated ∆E_ads values, however, can suffer from the basis set superposition error (BSSE). Indeed, upon correction, results indicate that these two complexes are nearly degenerate (BSSE-corrected ∆E_ads values being −9.4 and −8.9 kcal mol⁻¹ for the BN/NH₂ and BN/π, respectively).

The interaction of Gly through the COOH group by means of a simultaneous CO–B dative bond and a OH···N_(BNNT) H-bond has also been considered (see BN/COOH-1 and BN/COOH-2 adducts). Although none of the calculated complexes are the most stable ones, important structural and energetic features deserve to be mentioned. For the BN/COOH-1 adduct, Gly adsorption on the (4,0) BNNT results in a spontaneous proton transfer from the Gly COOH group to the N nanotube atom, hence forming a COO-/BNNT-H⁺ ion pair (see Figure 1b, structure of right). Such a proton transfer was already observed for the adsorption of HCOOH on the very same (4,0) BNNT and is attributed to the net charge transfer occurring from Gly to the BNNT, which induces an increase of the COOH acidity and the nanotube basicity up to the point of promoting the proton transfer to a nearby N atom of the nanotube. Moreover, for this adduct the CO–B distance is significantly shorter than for the other BN/COOH-1 adducts (1.487 Å versus 2.514–2.921 Å, respectively), which results in a stronger interaction (∆E_ads = −31.7 kcal mol⁻¹ and ≈ −9.6–−7.5 kcal mol⁻¹, respectively). For the BN/COOH-2 complex on the (4,0) BNNT, no proton transfer has been found; although the OH···N_(BNNT) H-bond and the CO–B dative bond are actually shorter than those present in the other BNNTs, in line with what has been described for the BN/COOH-1 cases. Interestingly, the difference between BN/COOH-1 and BN/COOH-2 is that in the former the COOH interaction occurs on B and N atoms that are chemically bonded to each other, whereas in the latter this is not the case. Accordingly, the fact that the spontaneous proton transfer only occurs in the former system seems to indicate that the charge transfer in enhanced by a cooperative effect between the OH···N_(BNNT) and the CO–B interactions when the B and the N atoms are chemically bonded, which is in agreement with the larger and more negative calculated ∆E_ads values (−31.7 and −17.6 kcal mol⁻¹ for BN/COOH-1 and BN/COOH-2, respectively).

Finally, it is worth mentioning that the interaction of Gly in its zwitterionic state has also been computed (BN/zwitt). On the (4,0) and (6,0) BNNTs, a spontaneous proton transfer from the NH₃⁺ group to the N nanotube atom has been found, whereas for the (9,0) and (15,0) BNNTs the zwitterionic form is maintained. Consistently, calculated ∆E_ads values are negative for the two former adducts (−31.2 and −13.5 kcal mol⁻¹, respectively), whereas for the two latter ones they have been found to be positive (+1.5 and +7.6 kcal mol⁻¹, respectively) and, accordingly, are not stable complexes.

Here, results on the interaction of Gly with the BNNTs in the presence of seven water molecules are reported. We have introduced seven water molecules since this is the minimum number of water molecules to have a relatively complete first-solvation shell of Gly upon adsorption; i.e., three water molecules interacting with the NH₃⁺ group, two water molecules interacting with the COO⁻ group and two more water molecules to complete the solvation shell. For these cases, the unit cell parameters of the BNNTs have been enlarged thrice to avoid lateral interactions between water molecules of adjacent unit cells. It is worth mentioning that a statistical sampling of the hypersurface of these systems can be carried out adopting either the molecular dynamics or the Monte Carlo approaches [41]. However, these calculations are extremely expensive at the ab-initio level adopting realistic models for BNNTs like the (9,0) and the (15,0) ones, which contain 129 and 201 atoms, respectively. For the present work, we have followed a different approach consisting of a progressive microsolvation of the dry interface at the Gly/BNNT structures. This microsolvation procedure consists of adding water molecules at the dry Gly/BNNT interface in such a way that Gly progressively loses direct contact with the BNNTs up to a point in which the interaction is fully bridged by water. This procedure has already been performed by some of us in other works for the interaction of Gly with silica [42] and hydroxyapatite [43] surfaces. Since the most stable state of Gly in water is the zwitterionic one, we considered the BN/zwitt adducts as initial guesses for the progressive microsolvation. For the (4,0), (6,0) and (9,0) BNNTs, the resulting structures are shown in Figure 2a. In the BN/CONH adduct, Gly directly interacts with the BNNTs in a similar fashion as in the gas-phase (i.e., COO⁻···B_BNNT dative bond and NH₃⁺···N_BNNT H-bond), while the seven microsolvating water molecules are simple spectators interacting with available points of the COO⁻ and NH₃⁺ groups through H-bonding. It is worth remarking that now for the (4,0) and (6,0) BNNTs no H transfer from Gly to the BNNT occurs (at variance with the gas-phase adsorption, vide supra) due to the screening effect of water. The BN/CO-w/NH and BN/NH-w/CO adducts result from moving one spectator water molecule from the outer shell to the inner shell so that the following water mediated interactions NH₃⁺···H₂O···N_BNNT and COO⁻···H₂O···B_BNNT occur respectively. Finally, From these two adducts, a second water displacement to remove the remaining direct Gly/BNNT interaction gives the w/CONH adduct, in which water fully mediates the Gly/BNNT contact. It is worth mentioning that each of the seven H₂O molecules can in principle be displaced from their positions to lead to a water mediated contact between Gly and BNNT. We choose the one exhibiting the weakest interaction energy with the other water molecules by computing the cost to remove one water molecule from the BN/CONH adduct by a single point energy evaluation for each H₂O.

For the (15,0) BNNT, all the optimization calculations collapsed to structures with no direct contact between Gly and BNNT, the most stable one being presented in Figure 2b. In this structure, water fully solvates the Gly molecule and, at variance with the other BNNT cases, no charge transfer between water and BNNT takes place, due to the highly apolar character of this nanotube.

The relative stabilities between the different calculated adducts for a given microsolvated complex are shown in Table 3. As one can observe, for the (6,0), (9,0) and (15,0) BNNTs the most stable systems are the w/CONH adducts; i.e., those in which no direct Gly/BNNT contact occurs, whereas for the (4,0) one, the BN/CO-w/NH adduct (direct interaction only through the COO⁻) was found to be the most energetically stable one. It is worth mentioning, however, that the energy difference between the BN-CO-w/NH and w/CONH adducts for the (4,0) BNNT (the first and second most stable ones) is relatively small (2.6 kcal mol⁻¹) and, accordingly, it might be inverted due to entropic effects associated with water rearrangement, as it is shown for peptide adsorption on hydrophobic and polar surfaces [44]. To further analyze this point, finite temperature molecular dynamics simulations would be desirable.

Besides these results, three different processes have moreover been considered to study the stability of the structures shown in Figure 2. The first one involves Gly solvated by 7 H₂O molecules being adsorbed on the clean walls of the BNNTs, whose reaction energy was computed as (reported by ∆E_R1 of Table 2) Gly/7w + BNNT → Gly/7w/BNNT (2) where Gly/7w is glycine solvated by the seven H₂O molecules and Gly/7w/BNNT represents the microsolvated complexes. The ∆E_R1 column shows that for all BNNTs the process is exoenergetic, meaning that the structural rearrangement of the seven H₂O molecules around Gly is compensated by the interaction with the BNNTs. Remarkably, limited to the most favorable adducts per BNNT, ∆E_R1 values are less negative with increases in the nanotube radius, consistent with the less polar behavior of the BNNTs.

The second process envisages gas phase Gly adsorbed on the already microsolvated BNNTs by the seven water molecules, whereas the third one envisages the solvated Gly being adsorbed on the seven water hydrated BNNTs giving rise to the shown adducts with expulsion of seven water molecules (here considered as a H-bonded cluster), whose reaction energies were computed as (reported by ∆E_R2 and ∆E_R3 of Table 2) Gly + 7w/BNNT → Gly/7w/BNNT (3) Gly/7w + 7w/BNNT → Gly/7w/BNNT + 7w (4) where 7w/BNNT is the BNNT solvated by the seven H₂O molecules and 7w the H-bonded cluster made up by seven water molecules. These last two processes require the BNNTs in the presence of seven water molecules, whose optimized structures are given in Figure 3. The initial guess of these systems were the corresponding w/CONH adducts (no direct interaction between Gly and the BNNTs and accordingly the interaction between water molecules and BNNTs is maximum) in which the Gly molecule was removed. Consistent with the polar character of the BNNTs, several interactions between the water molecules and the (4,0) BNNT via covalent dative and H-bond interactions take place, whereas by increasing the BNNT radius these interactions are progressively missed, up to the point in which for the (15,0) no apparent interaction is observed.

The energetics of R2 is a tradeoff between the water affinity of the BNNTs and Gly (R2 is in essence the capture of the adsorbed water by Gly). The trend provided by the calculated ∆E_R2 values indicates that R2 is more favorable when increasing the radius. That is, since the (15,0) BNNT does not exhibit water affinity, the calculated ∆E_R2 values are large and negative due to the strong interaction of water with Gly. In contrast, both the (4,0) BNNT and Gly exhibit large water affinity and accordingly the calculated ∆E_R2 values are the less negative ones of the series. R3 is probably the most physically sound process as it involves the replacement of adsorbed water solvent by an already solvated Gly. Calculated ∆E_R3 values indicate that such a water replacement is energetically favorable on the (15,0) and (9,0) BNNTs probably due to the low water affinity of these BNNTs. On the (6,0) the process is still favorable by some amount but less than on the other two nanotubes (∆E_R3 = −2.5 kcal mol⁻¹), whereas on the (4,0) calculated ∆E_R1 was found to be positive, indicating that the overall interactions between Gly, water solvent and the nanotube are not as stable as the interaction between this nanotube and water.