Highly branched and loop-rich gels via formation of metal–organic cages linked by polymers

Gels formed via metal–ligand coordination typically have very low branch functionality, f, as they consist of ∼2–3 polymer chains linked to single metal ions that serve as junctions. Thus, these materials are very soft and unable to withstand network defects such as dangling ends and loops. We report here a new class of gels assembled from polymeric ligands and metal-organic cages (MOCs) as junctions. The resulting ‘polyMOC’ gels are precisely tunable and may feature increased branch functionality. We show two examples of such polyMOCs: a gel with a low f based on a M2L4 paddlewheel cluster junction and a compositionally isomeric one of higher f based on a M12L24 cage. The latter features large shear moduli, but also a very large number of elastically inactive loop defects that we subsequently exchanged for functional ligands, with no impact on the gel's shear modulus. Such a ligand substitution is not possible in gels of low f, including the M2L4-based polyMOC.

All deuterated solvents were purchased from Cambridge Isotope Laboratories, Inc. All other reagents and solvents were purchased from VWR® or Sigma-Aldrich®. All purchased reagents and solvents were used as supplied unless otherwise noted. All air-sensitive reactions were executed using standard Schlenk techniques. All filtration in vacuo, unless stated otherwise, was carried out over MAGNA nylon filter disks (Maine Manufacturing, LLC) with 0.45 µm pore size. Spectra/Por® 7 standard regenerated cellulose dialysis tubing (8 kDa molecular weight cutoff (MWCO), 25.5 mm diameter) was purchased from Spectrum® Laboratories.

Chromatography methods
Liquid chromatography-mass spectrometry (LC/MS) were performed on an Agilent 1260 LC system equipped with an Advanced Materials Technology HALO® C18 high performance column. Solvent gradients consisted of mixtures of Milli-Q® water with 0.1% acetic acid SUPPLEMENTARY INFORMATION DOI: 10.1038/NCHEM.2390 4 (AcOH) and HPLC-grade acetonitrile. Mass spectra were obtained using an Agilent 6130 single quadrupole mass spectrometer.
Preparative high performance liquid chromatography (prep-HPLC) was performed on an Agilent Technologies 1260 Infinity system equipped with a ZORBAX 300SB-C18 PrepHT column (ID x Length = 21.2 x 150 mm; particle size = 5 µm). Eluent flow rate was 20 mL/min, and the eluent composition consisted of mixtures of nano-pure water with 0.1% acetic acid (AcOH) and HPLC-grade acetonitrile. The eluent gradient consisted of a linear ramp from 20% to 60% acetonitrile during 0-18 min, followed by a ramp to 100% acetonitrile during 18-20 min. Polymer samples were dissolved at a concentration of 100 mg/mL, and injected in 1.0 or 0.50 mL volumes. The instrument was controlled using the OpenLAB PrepLC software.
Column chromatography was performed on a Biotage® Isolera One with Accelerated Chromatographic Isolation TM flash chromatography system, using Biotage® KP-Sil SNAP cartridges at the recommended flow rates (e.g., 50 mL/min for 100 g SNAP cartridge).
Gel permeation chromatography (GPC) measurements were performed in tetrahydrofuran (THF) using an Agilent 1260 Infinity system with variable-wavelength diode array (254, 450, and 530 nm) and a refractive index detector, guard column (Agilent PLgel; 5µm; 50 x 7.5 mm), and three analytical columns (Agilent PLgel; 5µm; 300 x 7.5 mm; 10 5 , 10 4 , and 10 3 Å pore sizes). The instrument was calibrated with low-dispersity polystyrene (PS) standards between 1.7 and 3150 kg/mol. All runs were performed at 1.0 mL/min flow rate at 25 °C . The numberaverage molar mass (M n ), weight-average molar mass (M w ), and dispersity index (Đ = M w /M n ) of PL1 and PL2 were calculated by applying the conversion from polystyrene samples described by Sadao and Mori (values of t = 0.916, s = 1.21, derived for PEG, were used for PL1 and PL2). S1 Solution nuclear magnetic resonance spectroscopy methods 1 H nuclear magnetic resonance ( 1 H NMR) and 13 C nuclear magnetic resonance ( 13 C NMR) spectra were recorded on two Bruker AVANCE-400 NMR spectrometers (NIH Grant # 1S10RR013886-01). Chemical shifts are expressed in parts per million (ppm), and splitting patterns are designated as s (singlet), d (doublet), t (triplet), m (multiplet), and b (broad). Scalar coupling constants J are reported in Hertz (Hz). MestReNova LITE v5.2.5-4119 software (Mestrelab Research S.L.) was used to analyze the NMR spectra. 1 H and 13 C NMR spectra were referenced to solvent peaks as reported in literature. S2 Magic angle spinning nuclear magnetic resonance spectroscopy methods Variable temperature 1 H magic-angle spinning solid-state nuclear magnetic resonance (VT 1 H MAS NMR) spectra were recorded on a 11.7 T (500 MHz, 1 H) home-built NMR spectrometer (courtesy of Dr. David Ruben, Francis Bitter Magnet Laboratory-MIT). The gel samples immediately after mixing components were loaded via syringe into a 4 mm RevNMRstyle zirconia rotor (60 µl fill-volume) which was sealed with a Kel-F cap to reduce the 1 H background signal. The spectra were collected using a spinning frequency (ω r/2 π) of 10 kHz with 128 co-added transients and a recycle delay of 3 seconds. Sample temperatures were varied between 20 and 70 °C and spectra were collected every 5 minutes over a period of eight hours. 1 H spectra were referenced to solvent peaks as reported in literature. S2
Matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF) was carried out using a Bruker Daltonics® Omniflex® MALDI-TOF instrument, operated using the FlexControl TM version 1.1 software. The data was analyzed using XMass software. The instrument was operated in a reflectron mode (200 ns pulsed ion extraction; reflector voltage = 19.7 kV; lens voltage = 9 kV; detector gain = 20x) with positive ion detection (ion source voltage = 19.0 kV). The laser was operated at a 60-65% intensity at a sampling rate of 1 shot per 1.0 ns, and ~200-1000 shots were averaged to achieve the desired signal-to-noise ratio. The target had a "Scout 49" geometry. The samples were prepared as follows: first, the matrix solution was prepared by making a saturated solution of α-cyano-4-hydroxycinnamic acid (CHCA) in 1:1 water/acetonitrile with 1% trifluoroacetic acid (TFA). Polymer samples were dissolved at a concentration of 2.5-5 mg/mL in acetonitrile. To 40 µL of the matrix solution was added 4 µL the solution of the polymer sample. 0.5 µL of this solution was spotted in several locations on the MALDI target plate. The reconstituted solution of Calibration Mixture 2 from the Sequazyme TM Peptide Mass Standards Kit (Applied Biosystems, part # P2-314300) was diluted 1:24 (v/v %) with the matrix solution, and 0.5 µL of this solution was spotted over half of the sample spots for each sample on the MALDI target plate. These standards provided for internal mass spectrum calibration. In cases where ionization of the polymer sample was hindered by the presence of the standards, external standard calibration was employed.

Microscopy methods
Cryogenic transmission electron microscopy (cryo-TEM) was performed on holey carbon grids by plunge freezing in liquid ethane using a Gatan Cryo-Plunge3 instrument, a Gatan Liquid Nitrogen Single Tilt Holder, and a JEOL 2100 FEG microscope. The samples were prepared similarly to the standard preparation of polyMOCs (vide infra), except at a low concentration ([PL1] = 4.6 mM) to afford soluble hyperbranched network fragments. Thus, PL1 (6.3 mg, 2.3 µmol) in a 2-mL scintillation vial was dissolved in 350 µL DMSO-d 6 , and to it was added a solution of Pd(NO 3 ) 2 . 2H 2 O (0.67 mg, 2.5 µmol) in 150 µL DMSO-d 6 via micropipette; the reaction mixture was heated at 70 °C for 1 d. 200 µL of the resulting solution was then dialyzed against Milli-Q® water (700 mL, 1 d) in a 8 kDa MWCO dialysis tubing (vide supra), affording 760 µL of aqueous solution, which was used directly (undiluted) for microscopy.

Molecular dynamics simulation methods
The simulated model system consisted of fully atomistic ligands and metal ions whose interactions were mediated by an implicit solvent. The effect of the implicit solvent is to (1)  Model. 1999, 5, 196-202] to describe Pd 2+ and the metal-ligand coordination. In the CaDA model metal ligand binding is described empirically through Coulombic interactions between partial charges on the ligand molecules and those on a model Pd 2+ complex consisting of a neutral Pd core bonded to four dummy atoms, each with a partial charge of +0.5 and arranged in a square planar geometry.

SUPPLEMENTARY INFORMATION
Simulations were carried out using GROMACS (version 5.02) [Berendsen, H.J.C., van der Spoel, D. and van Drunen, R., GROMACS: A message-passing parallel molecular dynamics implementation, Comp. Phys. Comm. 91 (1995), [43][44][45][46][47][48][49][50][51][52][53][54][55][56]. For the L1-type ligands the inter-and intra-molecular interactions, along with the Ligand-Pd 2+ Van der Waals interactions were described using a standard model force field. The details of the force field, along the associated GROMACS topology files, were adapted directly from simulations of Yoneya and Fujita, which can be found in the Methods sections of Ref. 51 of the main text. Therefore, our model of the para-substituted bis-pyridine (L1) monomers differ from that described in reference 51 only in that our model includes both the repulsive and attractive contribution to the short-range Lennard-Jones potential (in reference 51 they use a modified GROMACS package that includes the repulsive-only WCA short range potential). The additional presence of Lennard-Jones attractions in our simulations results in a decrease in overall computational efficiency, but is otherwise expected to have a negligible effect on the results. To model the L2-type ligand (the metasubstituted bis-pyridine) we modify the L1 model by exchanging the position of adjacent nitrogen and carbon atoms while retaining the force field parameters and partial charges of the original para-substituted ligand (see Supplementary Figure S27). Reorganizing atoms in this manner allows us to isolate the influence of the Pd-N bite angle on the the metal-ligand complexes.
In the gel-forming system each ligand is bound to a partner ligand via long and flexible polymer chains. An explicit model of such a system dramatically increases the number of atoms being simulated and also introduces a long timescale relaxation process associated with chain entanglement. Here we have utilized an efficient compromise in which the effect of the flexible polymer linker is described implicitly, in the form of a ligand-ligand pair potential. The pair potential is meant to mimic the effect of the polymer linker on the relative positions of connected pairs of ligands. The effect is primarily entropic and is, as we have treated it, directly related to the statistics that govern the end-to-end distance of the isolated polymer.
We modeled the effect of flexible polymer linker through the addition of a pair potential, , acting between the bridging carbon atoms of pairs of ligand molecules (those with partial charge of -0.162 in Fig. S27). In particular we took ( ) to be equal to the potential of mean force governing the end-to-end distance of a model polymer linker. This pair potential was generated based on simulations of an idealized version of the PEG linker. Given that the persistence length of PEG, = 3.8Å, S3 is approximately equal to the monomer size we modeled SUPPLEMENTARY INFORMATION DOI: 10.1038/NCHEM.2390 7 configurations of the 2.2kDa linker as a three-dimensional self-avoiding random walk with steps of length randomly distributed around a unit sphere. By sampling the statistics of this idealized model polymer we determined the end-to-end PMF and tabulated it for use as a user-input pair potential in GROMACS. Supplementary Figure S28 contains a plot of the potential of mean force used to model the 2.2kDa linker.
The potential used to describe the flexible ligand-ligand polymer linker imparts forces that are much smaller than those involved in ligand-metal coordination. The distribution of cluster sizes is therefore quite insensitive to the details of ( ). However, because this potential imposes spatial correlations on specific pairs of ligands, the network-forming properties system can be sensitive to the specific details of ( ). The sensitivity is most pronounced in the statistics of inter-MOC connectivity where the probability for loop formation can be controlled by varying the shape of ( ). For instance, can be chosen to mimic a very short linker (keeping ligand pairs very close) so that loop formation is enhanced or to mimic a long linker (leveraging the entropic driving force that prevents configurations with very small end-to-end distances) so that loop formation is reduced. Our procedure is approximate but gives rise to ligand-ligand correlations that are in reasonable agreement with our physical expectations.
Simulations consisted of 96 metal ions and 192 ligands (or 96 macromers), enough to form four fully assembled M 12 L 24 type clusters, in a cubic, periodically replicated simulation cell with side length 18.7nm. Initial configurations were generated by randomly distributing ligands and metal ions subject to the constraint that the initial separation between any two species be greater than 1.5nm. For the polyMOC formation, PL1 and PL2 were placed randomly but at an initial fixed distance of 2.6 nm for the 2.2kDa PEG chain. Simulations were carried out in the NVT ensemble with a 2 fs time step. For each ligand model (4 in total, i.e., L-para, L-meta, and PL1 and PL2) we generated 20 individual trajectories. Each trajectory consisted of an initial 1 ns equilibration run at a temperature of 500K, followed by a 1 production run at a temperature of 350 K. The details of the implicit solvent and the thermostat were identical to those described in Ref. 52.
The results presented in Supplementary Figure S26A of the main text indicate that the Lpara ligands tend to form large and sometimes system-spanning clusters. To explore the effect of concentration on the formation of very large ligand-metal clusters we carried out a set of simulations in which 192 L-para ligands and 96 metal ions were placed in a larger periodically replicated cubic cell, one with side length 30 nm. In Supplementary Figure S26B we present the distribution of cluster sizes, ( ), that emerged from this 'low concentration' simulation compared to that of the high concentration simulation (the latter is an excerpt of the data plotted in Supplementary Figure S26A). To facilitate a side-by-side comparison, each curve has been normalized over the same increment, specifically 0 ≤ ≤ 50. At low concentration we observe a significantly reduced probability for observing clusters with > 50, and in fact the distribution of cluster sizes for ≤ 50 are quite similar for the low and high concentration simulations. This indicates that the presence of large clusters is either the result of exceeding a percolation threshold in ligand concentration or a preferential stabilization of such clusters due to selfinteraction through the periodic boundaries of the system.

Crystallography methods
Low-temperature (100 K) diffraction data (φ-and ω-scans) were collected on a Bruker X8 Kappa Duo four-circle diffractometer coupled to a Smart Apex2 CCD detector, with Mo K α SUPPLEMENTARY INFORMATION DOI: 10.1038/NCHEM.2390 8 radiation (λ = 0.71073 Å) from an IµS micro-source. The diffractometer was purchased with the help of funding from the National Science Foundation (NSF) under Grant Number CHE-0946721. The structure was solved by direct methods using SHELXS S4 and refined against F 2 on all data by full-matrix least squares with SHELXL-97 S5 following established refinement strategies S6 .
The final cif file was checked using the IUCr checkCIF routine, and below, we list the Alerts of level A and B as they appear in the output checkCIF file and the justification for each.

Alert level A
SHFSU01_ALERT_2_A The absolute value of parameter shift to su ratio > 0.20 Absolute value of the parameter shift to su ratio given 6.596 Additional refinement cycles may be required. Author Response: Structure refinement is not complete. Data are of low quality and so is the structure. Paddlewheel connectivity is confirmed but not much else can be concluded from this structure. Author Response: Structure refinement is not complete. Data are of low quality and so is the structure. Paddlewheel connectivity is confirmed but not much else can be concluded from this structure.

PLAT601_ALERT_2_B Structure Contains Solvent Accessible VOIDS of . 192 Ang3
Author Response: Structure refinement is not complete. Data are of low quality and so is the structure. Paddlewheel connectivity is confirmed but not much else can be concluded from this structure.
PLAT934_ALERT_3_B Number of (Iobs-Icalc)/SigmaW > 10 Outliers .... 3 check Author Response: Structure refinement is not complete. Data are of low quality and so is the structure. Paddlewheel connectivity is confirmed but not much else can be concluded from this structure.

Small-angle neutron scattering (SANS) methods
Small-angle neutron scattering (SANS) measurements were performed at the National Institute of Standards and Technology (NIST) Center for Neutron Research (NCNR) (Gaithersburg, MD, USA). The scattered neutron intensity was measured as a function of scattering variable q, where q =(4π/λ) sin (θ/2) and θ is the scattering angle. The beam was monochromated to a wavelength, λ, of 6 Å. Three sample-to-detector distances of 1 m, 4 m, and 13 m were used to cover a total q range of 0.004 to 0.5 Å -1 . 400 µL of each sample (gels prepared at 3.54 wt. % of polymer network) was loaded into titanium sample cells with a 1 mm path length. Experiments were performed on the NGB 30 m SANS instrument. Collected data were reduced and analyzed using the SANS macros package provided by the NCNR S7 . The resulting data were placed on an absolute scale and corrected for background electronic noise, detector inhomogeneity, and empty cell scattering using standard techniques.
Scattering data for both samples were fit using a sum of two models, the power law model and the core chain model. The power law model is primarily used to show the presence of a larger entangled network, and describes the scattering intensity as I(q) = Aq -n . The core-chain model, due to Hore et al. S8 , is a slightly modified version of the reported model. S9 The original Hore core-shell-chain model in ref. 2 was used to describe an inorganic iron oxide core with a shell layer of dense polymer brush, surrounded by grafted polymer chains with excluded volume. Here, the shell layer element is omitted but the model remains intact otherwise. In addition, the core in the present system is not inorganic entirely, but a mixed composition of Pd and bispyridine ligand. Most importantly, the model does not assume that the chains are Gaussian, and allows the excluded volume of the chains to vary. For this reason, Debye functions are omitted in favor of a more detailed description of polymer chain scattering.
The scattering intensity is calculated from the sum of the spherical core form factor, corechain form factor correlations, chain-chain correlations, and the form factor of a polymer chain with excluded volume. The form factor amplitude of the spherical core is given by F A (q), where j 1 is a spherical Bessel function of order 1, r core is the radius of the paddlewheel or cage, V core is the volume of the paddlewheel or cage, ρ core is the scattering length density (SLD) of the paddlewheel or cage, and ρ solvent is the SLD of the solvent.
Scattering from polymer chains is described by the form factor amplitude and form factor of the polymer chains, F B (q) and P B (q), respectively. Note that because polymer chains are fractal in nature, the form factor is a separate function from the form factor amplitude. The functions are given by, where the lower incomplete gamma function reads The parameter U = q 2 a 2 N 2 ν /6 contains the scattering variable q, the statistical segment length of the polymer chain (a), the degree of polymerization of the chain (N), and the excluded volume parameter ν . The total macroscopic scattering cross section for N p /V density of nanoparticles with N g grafted polymer chains per particle, including the power law term, is then expressed as where E A = j 0 (qr core ) is a spherical Bessel function, and B is the constant incoherent background. The radius of gyration for the chains surrounding the paddlewheel or cage is calculated from the parameters of Eq. (1.5) as ( 0.5) The SLDs for the core of each sample were calculated using an average of the PEG and palladium SLD on the basis of the composition of the two components. Using the NCNR SLD calculator, the bis-pyridine SLD was 1.98 x 10 -6 / Å 2 , the Pd SLD was 4.02 x 10 -6 / Å 2 , and DMSO-d6 had an SLD of 5.28 x 10 -6 / Å 2 . The composition of the components, calculated on the basis of the density and mass for paddlewheel and cage gels (gel-2 and gel-1, respectively), was 70% ligand and 30% Pd yielding an initial SLD of 2.59 x 10 -6 / Å 2 . The initial SLD does not take into account the possible presence of PEG or DMSO within the core, and so is only an initial approximation. The calculated radii, resulting from the core-chain model fits, for the paddlewheel and cage structures were 0.55 ± 0.054 nm and 1.70 ± 0.25nm, respectively. The calculated number of ligands, also from the core-chain model fits, for the paddlewheel and cage structures, N g , were approximately 4 and 20, respectively. The excluded volume parameter (ν ), calculated from the core-chain model fit, for the paddlewheel gel and cage gel, were 0.574 and 0.595, respectively. A value of ν that is close to 0.6 is indicative of a swollen polymer chain (i.e., R g ~ N 0.6 ). The radius of gyration (R g ) calculated using Eq (1.6) with paramaters obtained from the core-chain model fits were 0.49 nm and 0.45 nm, respectively, for paddlewheel and cage gel.
We acknowledge the support of the National Institute of Standards and Technology, U.S. Department of Commerce, in providing the neutron research facilities used in this work. This work utilized facilities supported in part by the National Science Foundation under Agreement No. DMR-0944772. This manuscript was prepared under cooperative agreement 70NANB12H239 from NIST, U.S. Department of Commerce. The statements, findings, conclusions, and recommendations are those of the authors and do not necessarily reflect the view of NIST or the U.S. Department of Commerce.

DOI: 10.1038/NCHEM.2390
Frequency sweep and strain sweep experiments were performed on an Anton Paar MCR 301 rheometer. The rheometer was outfitted with a Peltier heating system with an environmental enclosure for temperature control. A disposable parallel-plate geometry (radius = 12 mm) was used and coupled with a disposable bottom plate, with the typical gap of 1.00 mm between the two plates. Frequency sweep experiments were performed from 0.1 to 100 rad/s at 1% strain, which was first confirmed to be in the linear viscoelastic regime using strain sweep experiments. Gel samples were prepared either on the plate in-situ or in 1-dram vials (vide infra) and subsequently transferred onto the rheometer. Experiments were performed at 25 °C and the evaporation of solvent (DMSO-d 6 ) was negligible within the typical measurement time (< 1 hour).

Fluorimetry methods
Fluorimetry was carried out using Fluorolog®-3 spectrofluorometer from Jobin Yvon Horiba using the DataMax for Windows TM driving software. The following parameters were used during the fluorimetry: (1) integration time = 0.25 s; (2) increment = 1 nm; (3) excitation wavelength: 340 nm; (4) Detector HV S = 950 V and R = 0 V; bandpass slits: excitation1 = 3.000 nm; emission1 = 5.000 nm. The data was analyzed using OMNIC TM software and plotted in OriginPro 8.5. The samples for fluorimetry were prepared by depositing a fragment of the gel into a cylindrical hole (2 mm diameter x 0.9 mm depth) within a sample holder, sandwiching the sample holder with the gel between two square glass cover slips to stabilize the gel and prevent solvent evaporation, and using the front-face geometry to collect the fluorescence.

Gel swelling methods
Two sets of gels (3 gels per set) were prepared in tared 1-dram vials -one set from macromer PL1 and the other set from macromer PL2 -following the general polyMOC synthesis method (vide infra), except at 0.500 x the scale: i.e., 10.13 mg of macromer was used to prepare gels with [macromer] = 24 mM (Supplementary Figure S31). To each gel was added 3.8 mL DMSO, and the gels were allowed to stand at RT for 5 d, at which time the excess DMSO was removed via syringe, and any residual DMSO was wicked away by gently dabbing the gels in the vials with Kimwipes®. Vial inversion tests confirmed the materials remained in the gel state (see Supplementary Figure S31). The gels in the vials were weighed, and the swelling ratio for each was determined by diving the mass of the swollen gel by the dry mass of the network (dry mass of the network = mass of macromer + mass of Pd(NO 3 ) 2 . 2H 2 O = 10.13 mg + 1.00 mg = 11.13 mg) (Supplementary Table S3). Averages and standard deviations were computed for the three trials.
Computations of network branch functionality f from G' via phantom network theory of rubber elasticity S10, S11 A measurement of G'~|G| (i.e., G' >> G'') allows us in principle to compute f in our polyMOCs and thereby validate the conclusions derived from the simulations. The phantom network theory was deemed most appropriate for the analysis of our polyMOCs, because it explicitly relates |G| and f: S11

SUPPLEMENTARY INFORMATION
where ρ is the mass density of the elastically active polymer chains, R is the universal gas constant (8.31446•10 6 cm 3 •Pa•K -1 •mol -1 ), T is the temperature (in this case 298.15 K), M chain is the number-average molecular weight (M n ) of the polymer chain separating the junctions (taken here to be equal to M n of the macromers = 2700 g/mol). S11 Crucially, ρ is not known a priori because an unknown fraction of the polymer chains may form primary loop defects (and therefore not contribute to ρ (dangling chain ends were not observed by NMR and can therefore be ignored). A key realization is that to a first-order approximation (i.e., accounting only for primary loops) ρ scales linearly with the true f of the network -in other words a deviation of f from f ideal by x % is expected to yield the same x % deviation of ρ from ρ ideal . Thus, we have and substitution of the expression for ρ into the expression for |G| yields In the case of polyMOCs derived from PL1 or PL2, where no macromer is replaced with free ligand L1 or L2, f ideal is taken to be 24 or 4, respectively, values expected for M 12 L 24 or M 2 L 4 cages, and consistent with at 1 µs computed above for the corresponding networks (21 ± 6 and 5.3 ± 0.7, vide supra). When a fraction r of macromer PL1 or PL2 is replaced with L1 or L2, respectively, new f ideal becomes equal to (1 -r) • (f ideal when r = 0) because the free ligands do not contribute to the branch functionality: i.e., f ideal = (1 -r) • 24 for gel-1, and f ideal = (1 -r) • 4 for gel-2.
G' and G'' at ω = 10 rad/s were measured for all polyMOCs with varying fractions r of macromer replaced with L1 or L2 (Supplementary Table S4). Thus, when G' >> G'', as is the case for all of our polyMOCs at ω = 10 rad/s, G' can be used in place of |G|. Note: for the comparison of f for gel-1 and gel-2 before and after annealing, G' at ω = 100 rad/s was utilized. A sample calculation is provided below for a gel-1 gel where r = 0.125. The volume of the gel is the added volume of DMSO-d 6 and the polymer network (the latter was estimated to be the same as that of PEG with M n = 2000, which was measured to be 0.8273 mL/g). 2H 2 O (6.67 mg, 0.02503 mmol) in 133.3 µL of DMSO-d 6 . The resultant mixture was agitated briefly to afford a light-yellow homogeneous solution; the head-space of the NMR tube was briefly purged with argon, and the NMR tube was sealed. The mixture in the NMR tube was heated at 70 °C for 8 h.  19, 149.23, 145.51, 138.51, 138.38, 134.98, 127.82, 126.02, 124.07, 62.38. FT-ICR-ESI HRMS: calcd. for C 36 H 36 Br 4 N 4 [M -3(NO 3 -)] 3+ , most abundant m/z = 441.4132; found, 441.4141. *The resonance corresponding to the ROH proton is extremely broad and overlapping with nearby peaks, preventing accurate integration.

PL1.
To a 20-mL microwave vial equipped with a magnetic stir-bar were added finely powdered L1 (0.300 g, 1.14 mmol), PEG-diacid (M n = 2.2 kDa, 0.7741 g, 0.35 mmol), EDC . HCl (0.4336 g, 2.26 mmol), and DMAP (0.0870 g, 0.712), and the vessel was brought into the glove box. To the vessel was added DCM (3.84 mL), ensuring that the solid was washed down the walls of the vial, and the vial was crimped and set to stir at RT. During the first several minutes, most of the solid dissolved, giving rise to a yellow solution. After 25 h, the vessel was removed from the glove box and concentrated via rotary evaporation. The product was extracted from the oily mixture with toluene (3 x 2 mL), filtering the extracts through Celite® 545 packed in Pasteur pipettes. The combined extracts were concentrated by rotary evaporation, redissolved in 2 mL toluene, and precipitated into cold (~ -10 --20 °C) diethyl ether (35 mL). The vial containing the solution of crude product was rinsed with an additional 0.5 mL toluene, and this rinsing was also subjected to precipitation. The precipitation was allowed to proceed overnight at -20 °C, and the precipitated product was collected by filtration in vacuo, rinsing with cold diethyl ether (2 x 40 mL, 1 x 100 mL). Quickly, while still cold, the white solid was transferred from the filter disk to a 20-mL scintillation vial and was dried in vacuo overnight. The dry solid (~950 mg) was redissolved in water (9.5 mL, 18.2 MΩ . cm at RT, from Milli-Q® system), filtered through a nylon syringe filter with 0.45 µm-pore size, and subjected to prep-HPLC purification. The combined pure fractions were diluted with sat. NaHCO 3(aq.) until pH was ~7-8. The aqueous phase was divided into two portions and the product was extracted from each one with DCM (5 x SUPPLEMENTARY INFORMATION DOI: 10.1038/NCHEM.2390 16 300 mL). The combined extracts were dried over anhydrous Na 2 SO 4 , concentrated by rotary evaporation, re-dissolved in 1 mL DCM, and precipitated into cold diethyl ether (35 mL). The vial containing the dissolved pure product was rinsed with an additional 0.5 mL DCM, and this rinsing was also subjected to precipitation. The precipitation was allowed to proceed overnight at -20 °C, and the precipitated product was collected by filtration in vacuo, rinsing with cold diethyl ether (2 x 40 mL, 1 x 100 mL). Quickly, while still cold, the white solid was transferred from the filter disk to a 20-mL scintillation vial and was dried in vacuo overnight, affording PL1 (0.3752 g, 40 % yield) as a soft white solid. 1 H NMR (400 MHz, CD 2 Cl 2 , 25 °C): δ 8.69 (bdd,J = 4.3,1.4 Hz,8H),7.88 (t,J = 1.7 Hz,2H),7.73 (d,J = 1.7 Hz,4H),7.59 (dd,J = 4.5,1.7 Hz,8H),5.33 (s,4H),4.22 (s,4H),204H). 13 C NMR (100 MHz, CD 2 Cl 2 , 25 °C): δ 170. 55, 150.77, 147.64, 140.04, 138.23, 127.83, 126.09, 122.07, 71.33, 70.90 (b), 70.86, 68.98, 66.19 PolyMOC gels with macromer replaced with free ligand. General synthesis: The procedure is identical to that used for the synthesis of regular polyMOC gels, except during the preparation of the macromer solution, x% of the macromer was replaced with 2 equivalents of the corresponding free ligand to achieve the same total concentration of dipyridine ligands (e.g., when x% = 25%, instead of 20.25 mg of macromer PL1, 15.18 mg of PL1 (0.56 µmol) was combined with 0.98 mg of L1 (0.38 µmol) in 210.0 µL DMSO-d 6 ).

S1.
Mori, S. Calibration of size exclusion chromatography columns for determination of polymer molecular weight distribution. Analytical Chemistry 53, 1813-1818 (1981   A. Distribution of cluster sizes for systems with L-para and L-meta ligands are plotted with blue or red bars, respectively. The data representing polyMOC networks derived from PL1 and PL2 are plotted with points using the same colors. B. The probability distribution for observing a cluster of size in a low concentration simulation (system sidelength = 30 nm) is plotted with green boxes. For comparison, the similar distribution for high concentrations (system sidelength = 18.7 nm) is plotted in blue boxes, however scaled so that the normalization is equivalent on the increment of 0 ≤ ≤ 50.

Supplementary Figure S27
A rendering of the Pd 2+ complex, the para-substituted bis-pyridine ligand (L-para), and the meta-substituted bis-pyridine ligand (L-meta). The atoms are labeled with their respective partial charges, given in units of electronic charge.