TY - JOUR
T1 - Correction
T2 - CO2 induced phase transitions in diamine-appended metal-organic frameworks (Chemical Science (2015) 6 (5177-5185) DOI: 10.1039/c5sc01828e)
AU - Vlaisavljevich, Bess
AU - Schnell, Sondre K.
AU - Dzubak, Allison L.
AU - Lee, Kyuho
AU - Planas, Nora
AU - Neaton, Jeffrey B.
AU - Gagliardi, Laura
AU - Smit, Berend
PY - 2019/1/1
Y1 - 2019/1/1
N2 - The authors regret that there are some discrepancies reproducing the data in the original article due to the determined coordinates not being the fully optimised geometries. The authors have provided more information as follows. In the manuscript entitled 'CO2 induced phase transitions in diamine-appended metal-organic frameworks', minor errors with the attached coordinates and energies reported in the paper have recently been identified. In this communication, we correct these errors. Here, we present updated optimized geometries and binding energies. We also take this opportunity to include an extended computational details section to ensure reproducibility. In addition, we show that the overall conclusions of the paper are not affected by these changes. A detailed comparison with the results reported by Lee et al.1 revealed that the DFT optimization of the coordinates provided with the manuscript do not lead to the values reported in the manuscript, and they warrant correction. Corrected coordinates and updated tables (Tables 1-7) and figures (Fig. 1, 2, 4 and 5) are included here for calculations using the PBE functional. These structures have been repeated using a slightly tighter force threshold than in the original manuscript (details below). The M06-L calculations reported in the original manuscript are not revisited since they were performed to assess the role of dispersion. Since the publication of our work in 2015, a far more detailed study of this effect has been published by one of the authors rendering these M06-L calculations unnecessary and we refer readers interested in the role of dispersion on the carbamate formation to this more recent study by Lee et al.1 In addition to correcting our DFT calculations, we examine the effects of the revised DFT values on the lattice model in this work.We recompute the lattice model with the M06-L and PBE values fromthe original manuscript as well as the corrected PBE values reported below (Fig. 6-8 and Tables 8-10). In all three sets of isotherm plots the ordering is preserved but the inflection points are spaced differently with the new PBE numbers, leading to quantitative differences that are nonetheless qualitatively similar to previous work. Finally, we discuss different ways that CO2 can coordinate to the metal binding site, as shown in Fig. 3. We should have notedmore clearly in ourmanuscript that these were starting configurations and not necessarily the final converged structures since our goal was to try several starting geometries to determine which coordination environment around the metal site was lowest in energy. Take for example bidentate insertion. Chemical intuition suggests that this structure could rotate to one that has only one CO2 oxygen center closer to the metal than the other and we observe this in our optimized structure. The resulting geometries we obtained for the starting arrangements noted in the figure are higher in energy than the chain model as reported in our original paper.We wish to emphasize that at the time of our 2015 study, our objective was to understand whether or not CO2 was bound to the metal and if one-dimensional chain formation could lead to a step in the adsorption isotherm. It has since become clear that a far more thorough study of the arrangements of the amines is required to truly understand competing amine arrangements preset in experiment. This was outside the scope of our work. Once more, these calculations are perhaps now outdated given work in the field in recent years. We again refer interested readers to a more recent study by Lee et al.1 1. Extended computational details to ensure reproducibility In the course of rectifying the error in our calculations, we wanted to ensure that all revised calculations were converged using the exact same protocol; therefore, we repeated the PBE calculations for the pair and chain models using updated computational details given here to ensure reproducibility. The M2(dobpdc) MOF contains six unsaturated metal sites per unit cell. To calculate the binding energies of CO2 in its amine appended analogue mmen-M2(dobpdc), one mmen ligand per CO2 was added per unit cell. The smaller sized ethylenediamine (en) was used to saturate the remaining amines not involved in CO2 binding. In the case of the pair mode, two mmen-amines are included per unit cell only. All DFT calculations were performed with periodic boundary conditions carried out using the VASP 5.4.4 package (original calculations were performed with VASP 5.3.3). The PBE functional was employed to examine the energetics of CO2 adsorption.3 On-site Hubbard U corrections were employed for metal d electrons.4 The U values are determined to reproduce oxidation energies in the respective metal oxides and are given in the tables below. The electron-ion interactions in these calculations were described with the projector augmented wave (PAW) method developed by Blöchl with an energy cutoff of 550 eV.5 This combination of the PBE functional, PAW scheme, and energy cutoff was used for full geometry optimization of the various species investigated until the forces on all atoms were smaller than 0.02 eV Å-1 and the SCF convergence was set to 1 × 10-7 eV. Given the large size of the unit cell and the tests with other numbers of K-points from the original study, only results obtained from G-point calculations are reported here. Finally, heats of adsorption are now reported below along with E + ZPE values, while in the original manuscript only E + ZPE were reported. No changes were made to how the vibrational corrections were computed; however, we have included some additional details to ensure reproducibility.6 Harmonic vibrational modes (ωi) were computed for CO2 in the gas phase and its bound product state (amine-CO2-MOF complex). The framework itself was taken to be rigid and only the vibrational modes associated with the motion of the amine, the metal center, first coordination sphere (oxygen atoms bound to the metal in the MOF backbone), and (if present) the bound CO2 were computed. Since the harmonic approximation breaks down for low frequency modes, we replaced all modes less than 50 cm-1 with 50 cm-1 when computing the zero-point and thermal energies. The following standard harmonic expressions were used to compute the vibrational corrections: Zero-point vibrational energy (ZPE) is: [Equation presented here] While for the bound product, the rotational and translational degrees of freedom of CO2 have been converted to additional vibrational modes allowing one to compute the thermal correction simply as: [Equation presented here] 2. Values for the chain model The chain model used in our original study included 1 mmen- and 5 en-amines. The values from the original paper are reported in Table 1. When we repeat these calculations using the procedure described in Section 1, we obtain the values in Table 2. In addition to the chain model described above (1 mmen- and 5 en-amines per unit cell), during our original study we performed calculations with another model that was not included in the manuscript since its values yielded results further from experiment. This model includes only 1 mmen-amine per unit cell (no other amines) and was used to test the assumption that the five enamines are indeed spectators with respect to the metal dependence of the binding energy. We present the results from this model in Table 3. In the original paper we noted that the energy and bond length trends are correlated and are consistent with the Irving-Williams series. This is no longer true for all metals under investigation, with Zn being an outlier. The results for Zn can be explained by more recent work.1 3. Values for the pair model The model used to compute the "pair" adsorption mechanisms included 2 mmen-amines and 0 en-amines. The values in the original paper are presented in Table 5. 4. Lattice model plots The lattice models to generate adsorption isotherms for these systems were run at one temperature (∼25 °C) using four different input parameters. First the M06-L and PBE values from the original paper were used once more as it has been some time since we have run the lattice model. Then the model is repeated with the new set of values from PBE. If we compare Fig. 7 and 8, the order is preserved, but the infliction points are spaced a bit differently. This is due to the scaling factor being constant and is something we scaled for each of the different systems as well. The slope is also a bit different, but not more then we should expect for this simple lattice model. Furthermore, we only ever aimed to reproduce the step and the order of the metals. Any finer details cannot be expected to be obtained from this model. The exact values used to compute the isotherms are given in the tables below. The Royal Society of Chemistry apologises for these errors and any consequent inconvenience to authors and readers.
AB - The authors regret that there are some discrepancies reproducing the data in the original article due to the determined coordinates not being the fully optimised geometries. The authors have provided more information as follows. In the manuscript entitled 'CO2 induced phase transitions in diamine-appended metal-organic frameworks', minor errors with the attached coordinates and energies reported in the paper have recently been identified. In this communication, we correct these errors. Here, we present updated optimized geometries and binding energies. We also take this opportunity to include an extended computational details section to ensure reproducibility. In addition, we show that the overall conclusions of the paper are not affected by these changes. A detailed comparison with the results reported by Lee et al.1 revealed that the DFT optimization of the coordinates provided with the manuscript do not lead to the values reported in the manuscript, and they warrant correction. Corrected coordinates and updated tables (Tables 1-7) and figures (Fig. 1, 2, 4 and 5) are included here for calculations using the PBE functional. These structures have been repeated using a slightly tighter force threshold than in the original manuscript (details below). The M06-L calculations reported in the original manuscript are not revisited since they were performed to assess the role of dispersion. Since the publication of our work in 2015, a far more detailed study of this effect has been published by one of the authors rendering these M06-L calculations unnecessary and we refer readers interested in the role of dispersion on the carbamate formation to this more recent study by Lee et al.1 In addition to correcting our DFT calculations, we examine the effects of the revised DFT values on the lattice model in this work.We recompute the lattice model with the M06-L and PBE values fromthe original manuscript as well as the corrected PBE values reported below (Fig. 6-8 and Tables 8-10). In all three sets of isotherm plots the ordering is preserved but the inflection points are spaced differently with the new PBE numbers, leading to quantitative differences that are nonetheless qualitatively similar to previous work. Finally, we discuss different ways that CO2 can coordinate to the metal binding site, as shown in Fig. 3. We should have notedmore clearly in ourmanuscript that these were starting configurations and not necessarily the final converged structures since our goal was to try several starting geometries to determine which coordination environment around the metal site was lowest in energy. Take for example bidentate insertion. Chemical intuition suggests that this structure could rotate to one that has only one CO2 oxygen center closer to the metal than the other and we observe this in our optimized structure. The resulting geometries we obtained for the starting arrangements noted in the figure are higher in energy than the chain model as reported in our original paper.We wish to emphasize that at the time of our 2015 study, our objective was to understand whether or not CO2 was bound to the metal and if one-dimensional chain formation could lead to a step in the adsorption isotherm. It has since become clear that a far more thorough study of the arrangements of the amines is required to truly understand competing amine arrangements preset in experiment. This was outside the scope of our work. Once more, these calculations are perhaps now outdated given work in the field in recent years. We again refer interested readers to a more recent study by Lee et al.1 1. Extended computational details to ensure reproducibility In the course of rectifying the error in our calculations, we wanted to ensure that all revised calculations were converged using the exact same protocol; therefore, we repeated the PBE calculations for the pair and chain models using updated computational details given here to ensure reproducibility. The M2(dobpdc) MOF contains six unsaturated metal sites per unit cell. To calculate the binding energies of CO2 in its amine appended analogue mmen-M2(dobpdc), one mmen ligand per CO2 was added per unit cell. The smaller sized ethylenediamine (en) was used to saturate the remaining amines not involved in CO2 binding. In the case of the pair mode, two mmen-amines are included per unit cell only. All DFT calculations were performed with periodic boundary conditions carried out using the VASP 5.4.4 package (original calculations were performed with VASP 5.3.3). The PBE functional was employed to examine the energetics of CO2 adsorption.3 On-site Hubbard U corrections were employed for metal d electrons.4 The U values are determined to reproduce oxidation energies in the respective metal oxides and are given in the tables below. The electron-ion interactions in these calculations were described with the projector augmented wave (PAW) method developed by Blöchl with an energy cutoff of 550 eV.5 This combination of the PBE functional, PAW scheme, and energy cutoff was used for full geometry optimization of the various species investigated until the forces on all atoms were smaller than 0.02 eV Å-1 and the SCF convergence was set to 1 × 10-7 eV. Given the large size of the unit cell and the tests with other numbers of K-points from the original study, only results obtained from G-point calculations are reported here. Finally, heats of adsorption are now reported below along with E + ZPE values, while in the original manuscript only E + ZPE were reported. No changes were made to how the vibrational corrections were computed; however, we have included some additional details to ensure reproducibility.6 Harmonic vibrational modes (ωi) were computed for CO2 in the gas phase and its bound product state (amine-CO2-MOF complex). The framework itself was taken to be rigid and only the vibrational modes associated with the motion of the amine, the metal center, first coordination sphere (oxygen atoms bound to the metal in the MOF backbone), and (if present) the bound CO2 were computed. Since the harmonic approximation breaks down for low frequency modes, we replaced all modes less than 50 cm-1 with 50 cm-1 when computing the zero-point and thermal energies. The following standard harmonic expressions were used to compute the vibrational corrections: Zero-point vibrational energy (ZPE) is: [Equation presented here] While for the bound product, the rotational and translational degrees of freedom of CO2 have been converted to additional vibrational modes allowing one to compute the thermal correction simply as: [Equation presented here] 2. Values for the chain model The chain model used in our original study included 1 mmen- and 5 en-amines. The values from the original paper are reported in Table 1. When we repeat these calculations using the procedure described in Section 1, we obtain the values in Table 2. In addition to the chain model described above (1 mmen- and 5 en-amines per unit cell), during our original study we performed calculations with another model that was not included in the manuscript since its values yielded results further from experiment. This model includes only 1 mmen-amine per unit cell (no other amines) and was used to test the assumption that the five enamines are indeed spectators with respect to the metal dependence of the binding energy. We present the results from this model in Table 3. In the original paper we noted that the energy and bond length trends are correlated and are consistent with the Irving-Williams series. This is no longer true for all metals under investigation, with Zn being an outlier. The results for Zn can be explained by more recent work.1 3. Values for the pair model The model used to compute the "pair" adsorption mechanisms included 2 mmen-amines and 0 en-amines. The values in the original paper are presented in Table 5. 4. Lattice model plots The lattice models to generate adsorption isotherms for these systems were run at one temperature (∼25 °C) using four different input parameters. First the M06-L and PBE values from the original paper were used once more as it has been some time since we have run the lattice model. Then the model is repeated with the new set of values from PBE. If we compare Fig. 7 and 8, the order is preserved, but the infliction points are spaced a bit differently. This is due to the scaling factor being constant and is something we scaled for each of the different systems as well. The slope is also a bit different, but not more then we should expect for this simple lattice model. Furthermore, we only ever aimed to reproduce the step and the order of the metals. Any finer details cannot be expected to be obtained from this model. The exact values used to compute the isotherms are given in the tables below. The Royal Society of Chemistry apologises for these errors and any consequent inconvenience to authors and readers.
UR - http://www.scopus.com/inward/record.url?scp=85072290222&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85072290222&partnerID=8YFLogxK
U2 - 10.1039/c9sc90137j
DO - 10.1039/c9sc90137j
M3 - Comment/debate
C2 - 32133123
AN - SCOPUS:85072290222
VL - 10
SP - 8265
EP - 8272
JO - Chemical Science
JF - Chemical Science
SN - 2041-6520
IS - 35
ER -