Structural organization of space polymers

Space polymer amide was first posed by us theoretically¹ when a biological hydrophobic protein exhibited properties consistent with it being able to potentially form in gas phase space and we considered that its water of polymerization would also provide a source of bulk water for solvating further chemistry. Formation enthalpy calculations for polymer amide in the gas phase indicated that as each amide bond formed, the product was stabilized by the hydrogen bonding of its explicit water. We proposed in Ref. 1 that such a polymer could aid accretion in a molecular cloud or proto-planetary disk. We then devised experiments on asteroid material in chondritic Vigarano class alteration type 3 (CV3) meteorites to test these theories. Our top priority in the meteorite experiments was to prevent any terrestrial contamination. Every experiment had to have a minimum of steps in clean rooms, be conducted on small samples, in small containers, and be performed at room temperature to keep any extraterrestrial polymer from degrading. We knew that past investigators had ground up large samples via a pestle and mortar (very hard to clean), boiled the sample particles for 24 h, and then subjected them to numerous procedures for analysis—these experiments yielding amino acids but no intact polymer. Our careful approach using phase chemistry did yield results, and intact polymer was found to be present,² confirmed by extraterrestrial isotopes.³ The polymers discussed herein represent the highest peaks in the mass spectra of Ref. 3, and all consist essentially of the anti-parallel combination of two 11-residue glycine chains, with a length of about 5 nm. Asteroid particles on the micrometer scale, with a collectively large surface area, were placed, without pre-processing, in a Folch extraction vial (Fig. 1) containing chloroform as a dense lower hydrophobic phase and methanol/water as an upper hydrophilic phase. It is critical to our results that the polymers of interest were found to collect in the interphase between these solvents where conditions attract amphiphilic species of density near unity, such as the glycine/iron polymers studied here.

As to structures, the first observations of polymers of amino acids in meteoritic material^2,4 were followed by a report⁵ of a 4641 Da polymer in both Acfer 086 and Allende CV3 class meteorites. Subsequent higher resolution mass spectrometry³ revealed a core structure at 1494 Da that directly assembled via silicon bonding into three-legged triskelia³ observed at mass to charge ratios (m/z) 4641 and higher multiples of this. Mass spectrometry therefore suggests that the triskelia bond into extended sheets that, in principle, can cover any type of plane or curved surface via assembly into a tiled sheet of hexagons and pentagons. One example is the layer observed on the wall of spherical gas bubbles in liquid.³ Fibrous crystals from the same extract are here analyzed by x-ray diffraction to confirm many aspects of the 1494 Da polymer structure and its superassembly into macroscopic shapes.

A striking morphology to be seen at the interphase layer of the Folch extraction vial is that of the angled phantom-like tubes shown in Fig. 2. These appear less than 24 h after exposure of micrometer-sized meteorite particles to solvent extraction by chloroform/methanol/water as described previously^2,3,5 and in supplementary material Sec. S1 methods. The total volume of these is greater than that of the initial particles, and yet amino acid polymers comprise less than a few percent of the meteoritic contents. Here, it is conjectured that the observed rod-like 1638 Da subunits³ are able to form a very low-density space-filling lattice, with tetragonal linkage at silicon atoms. The internal space defined by this structure is filled with solvents under our experimental regime, but the polymer lattice without solvents has an estimated density of 32 mg cm⁻³, presented below. From its symmetry, the polymer lattice, unlike hydrogel, appears to be arranged in a regular crystalline mode. It is self-assembling in the presence of the ions originally present in the sample, which are mainly from minerals containing Fe, Si, and Mg. Regular internal channels direct the diffusion or convective flow⁶ of subunit polymers to the growing “edge” of the lattice where they are incorporated and removed from solution.

In this paper, we first discuss the x-ray diffraction data (Sec. II) and then focus in Secs. III, IV, and V on the structure underlying the shapes in Fig. 2 before presenting discussion and conclusions in Sec. VI.

Characterized by mass spectrometry as having a core mass of 1494 Da and comprising a double stranded anti-parallel poly-glycine of 11 residues each side that are closed out by iron atoms,³ this rod-like protein bonds in trimers (as triskelia) to form a continuous sheet.³ However, optical microscopy of a (disturbed) sample of the sheet material enveloping gas bubbles³ showed a tangled cluster of fibers. Following solvent extraction in a closed glass V-vial (Fig. 1), a solution of meteoritic extract (from Acfer 086) concentrated to dryness via slow evaporation over 1 month was rehydrated with the slight addition of water to develop the fiber-like crystals shown in Fig. 3. Three of these in a cluster were mounted on a nylon loop and subjected to x-ray diffraction at 0.979 Å on the APS Argonne synchrotron (methods in supplementary material Sec. S1).

Figure 4 shows a diffraction frame from one of these fibers. The x-ray data frames each represent integration over a 0.2° oscillation. The pattern remains constant across many tens of frames indicating a circular tube structure. Regardless of orientation, a constant central motif is seen, comprising short opposed arcs at a nominal first order spacing of 24.19 and 16.12 Å, with a much less intense additional arc at 12.06 Å and the hint of an arc at 9.64 Å. Apart from this fixed central motif, there were mostly full rings at 6.31 and 5.29 Å, with fainter rings at 4.80 and 3.52 Å, the 6.31 and 5.29 Å rings showing fourfold symmetry and a 45° orientation to the central motif.

The angle independence in the central motif suggested concentric round tubes with multiple diffracting layers, as illustrated in Fig. 5. In turn, this suggested that the tube walls could be gently curved quasi-planar lattices composed of the rod-like entities identified in the mass spectrometry³ of mass approximately 1600 Da and a typical length of about 5 nm. Several questions arise:

1. Which of the 1494, 1566, and 1638 Da polymers are in the crystal?

2. What is the symmetry of the lattice? Possibilities to consider are:

   • hexagonal, based upon the threefold vertex of a triskelion,³ which can be oriented in two ways relative to the axis of the tube;

   • rectangular, based upon a distortion of the triskelion vertex into a “T” pattern, in which 2:1 rectangular “bricks” can be stacked to form the mesh, again oriented in two ways parallel to the tube axis;

   • square, based upon a fourfold vertex. This presents a single orientation in relation to the tube axis.

3. Why are the outer rings set at 45° to the axis of the tube?

4. Are the tube walls layered in a simple spiral wind? The alternative would be concentric tubes, referred to as the “Russian doll” mode.

The two types of hexagonal tiling associated with a triskelion were ruled out by the lack of sixfold or threefold symmetry in the x-ray images. The rectangular “brickwork” tilings with T-shaped triskelion vertices do not generate the 45° rotation of the outer ring lobes relative to the central two-dimensional “ladder.” We therefore consider a square tiling with a fourfold (tetraskelion) vertex (Fig. 5) such as may exist when 1494 Da polymer rods (Fig. 6) are joined via silicon atoms [Fig. 7(a)]. Two possibilities for the vertex are illustrated in Fig. 7. Each has a square pattern of iron atoms set at 45° to the lattice rods, but only Fig. 7(a) fits the diffraction data in respect to the inter-vertex spacing and the diffraction angle from the iron grouping. Figure 7(b) is thought to be the most relevant one to the space-filling structure of Sec. III.

Square lattices were constructed using Spartan software^7,8 that includes Merck Molecular Force Field (MMFF) energy minimization and Q-Chem functionality. First, a single 1494 Da molecule was generated [Fig. 8(a)] in accord with the mass spectroscopy structure.¹ Each such rod was linked to three others by silicon atoms, with vertex details as shown in Fig. 7(a). Stacks of tetraskelia up to three layers deep were assembled and energy-minimized using MMFF, one example being shown in Fig. 9. The separate 1494 Da rods were found to be bound to each other along “edges” by hydrogen bonds with an energy of 71 kJ/mol totaled along the whole length of each contact edge. Attempts to stack the 1494 Da rods in the thin or “flat” direction did not generate hydrogen bonding, so this possibility was ruled out.

Measurements of the vertex-to-vertex distance on this simulated square lattice, after energy minimization via MMFF, gave a side h = 48.38 ± 0.2 Å (12 measurements). At each vertex in each layer, there are four Fe atoms that dominate the x-ray scattering; consequently, we test this calculated side h against the x-ray data, with results shown in Table I.

There is very good agreement between the modeled side of a square tiling of 1494 Da rods and all of the diffraction orders in the central motif. The observed central motif is a conserved ladder and not a “cross” because there is only one region of the tube wall that is always aligned to diffract in the direction perpendicular to the plane containing the tube axis.

On the supposition that the 1494 Da polymer is the structural unit of square tiling, the strongest diffraction signals will come from the four iron atoms at each tetraskelion vertex [Fig. 7(a)]. The four-lobed diffraction rings at 6.31 and 5.29 Å originate from the 45° planes with Fe atoms drawn in Fig. 7(a). These diffract to give relatively narrow rings because there is a depth of possibly several hundred layers in alignment through which these smaller square patterns extend. The calculated Fe spacing from MMFF modeling is 5.87 ± 0.11 Å [Fig. 7(a)], which lies between the principal 6.31 Å and lesser 5.29 Å readings in diffraction. We conjecture that there are two possible iron locations, causing a split in this ring. The space of minor variations is too large to easily explore in such a complex situation.

The MMFF assembly of Fig. 9 has a layer depth of 9.2 Å, but this spacing may (a) be too sensitive to tube curvature to give prominent diffraction or (b) be variable because there is not a radial constraint on the layers. When an additional lattice length h is added to expand the circumference by one lattice unit, the radius increase is h/2π = 7.6 Å, which is not a perfect match to the estimated layer spacing of 9.2 Å, tending to rule out concentric tube (“Russian doll”) packing.

In this set of x-ray diffraction images, there is no sign of the connecting glycine polymers, which is not surprising in view of (a) the much smaller diffraction of light elements relative to iron and (b) the open nature of the lattice, which does not constrain the polymer position.

The present x-ray data indicates that the small crystals of Fig. 3 are multiple-walled nanotubes composed of rolled-up sheets of a square lattice of 1494 Da polymers linked at fourfold vertices, which are a direct analog of the threefold vertices proposed for the 4641 Da polymer.³ 

Molecular clouds are regions within the high vacuum of space where there is a low level content of hydrogen, helium, and small molecules containing hydrogen, lithium, carbon, nitrogen, and oxygen,⁹ extending to ethylene glycol and poly-aromatic hydrocarbons.¹⁰ It is predicted that glycine, the simplest amino acid, should be able to form^1,11,12 but it has not been observed, possibly due to its polymerization that is exothermic¹ in the conditions of “warm, dense, molecular clouds.” Also present are very small amounts of heavier elements such as magnesium, silicon, and iron, providing a basis for the polymer amide observed,³ which is dominated by glycine and iron. In meteoritic material derived from asteroids, we have observed a number of glycine polymers in relatively simple mass spectra up to m/z 4641 and dominant rod-like polymers around m/z 1600.

If, in the near vacuum of a pre-solar molecular cloud or of the proto-solar disk, there had been a path to the formation of these polymers, using incoming glycine, iron, and water, then there would have been an advantage associated with presenting the largest possible cross-sectional area to intercept the most raw material. The polymer rod superassembly with the greatest capability to span volume would then be enhanced via exponential competition above less optimal ones. This structure would naturally have very low density.

The observation (Fig. 2) of phantom angled structures of total volume greater than the volume of meteoritic material initially added to a vial was an indicator that the components of a very low-density network were present in the meteorite before solvent extraction. In each of the Folch extracts of KABA, Acfer 086, and Allende, these components gravitated (upward) through the dense chloroform phase into the relatively thin interphase region, where they became sufficiently concentrated to re-assemble while floating as if in space. Isotopes at extraterrestrial levels^3,5 and the signature of a –C–C–N– amino acid polymer backbone in secondary ion mass spectrometry of the same Acfer 086 sample¹³ were proof that these polymer rod subunits (around m/z 1600) pre-existed in the meteorite and could not have formed in the vial.

With a known, small set of dominant polymers to consider (identified by mass spectrometry), we propose here that the 1638 Da polymer (observed at m/z 1639) forms the connecting rods of a largely empty space-filling lattice. The question we address is the likely nature of this lattice, given the materials available and the unique macro-scale morphology of angled shapes that has to be explained.

We will consider the creation of uniform repeating truly three-dimensional structures from a single type of rod of length h connected at either fourfold or sixfold vertices. The vertices are constrained to be of a single type in each of these cases, so that a single definite chemistry can apply throughout a structure. For example, the fourfold vertex can be based upon a group IV atom such as carbon, silicon, or germanium, with tetrahedral bond symmetry. Also, the connecting polymer rods will have end-for-end symmetry so as to form the lowest-entropy structure.

Just as the triskelion³ in two dimensions can most efficiently create an enveloping surface of hexagons (and occasional pentagons) using a minimum of polymer rods, so in the third dimension, we are led to consider flat hexagonally tiled sheets containing rods of length h connected perpendicularly by similar polymer rods that project “up” or “down” from neighboring vertices. One is immediately tempted to stretch this assembly vertically, so as to envelop more space with the same number of rods (Fig. 10, parts A—E, high vertices labeled “H,” low vertices labeled “L”). This puckers the hexagons so that their component rods make an angle α with the original plane [Fig. 10(a)]. Vertical stretching initially increases the volume of the structure per polymer rod but the effect fades as the hexagon area decreases, its effective side length (viewed from above) reducing as hcosα [Fig. 10(d)]. In terms of angle α, the volume of a quasi-cell defined at its ends by two of these distorted hexagons is

Setting $dV¯dα=0$⁠, we find that a maximum in $V¯$ exists at $sin α=1/3$⁠, i.e., α = 19.471°, when

The optimized volume for this stretched structure therefore occurs when the angle between one of the vertical connectors and any other rod at a vertex is the tetrahedral angle of (90° + α) = 109.471°. The structure with the optimized volume turns out to have the exact tetrahedral symmetry at each vertex. By following this approach, we have arrived at the 2H or hexagonal diamond structure¹⁴ that has a threefold rotational symmetry around the hexagon axis and mirror symmetry in the plane centered between the quasi-planes of the hexagonal mesh. The 2H unit cell is smaller than those of the 3C cubic diamond and higher polytypes such as 4H and 6H, so it is the most simple of all these structures.

In place of the carbon in a diamond structure, we propose that Si, another group IV element with tetrahedral bonding symmetry, is the atom at the position of each vertex in our space-filling structure of polymer rods. Silicon is plentiful in these meteorites and is available as a free ion in aqueous solution. The triskelion linker atom was identified as silicon in mass spectrometry,³ making Si our leading candidate for the structural linker between polymer rods in three dimensions.

For the connecting polymer rods in 3D, we invoke the 1638 Da polymer (seen at m/z 1639 and illustrated in Figs. 6 and 8) because it is available in quantity (this and the related 1566 Da polymer are dominant in the mass spectrum³) and it is end-for-end symmetrical, with an oxygen atom at each tip for bonding to Si. Effectively, a tetrahedral SiO₄⁴⁻ anion becomes the central entity at a vertex [Fig. 7(b)], with bonding to four surrounding Fe atoms. Each Fe atom is then linked to the pair of oxygen atoms at an end of a 1494 Da unit (Fig. 6).

An MMFF simulation^6,7 (at 298 K) gives the length of the 1638 Da rods as h = 4.806 nm, measured from vertex to vertex, to the central Si atom positions [Fig. 8(b)].

A measure of space-filling efficiency is the quotient Q = $V¯$/(number of rods per quasi-cell). Each quasi-cell contributes to the net four connecting rods because the six in the hexagon defining the end of a cell are each shared by a neighboring hexagon, making net three per cell and there is net one additional vertical connector per cell between the two hexagon ends. Consequently,

From (3), using the polymer rod, the connector mass of 1638 Da, plus the net ½ Si atom (14 Da), we derive a lattice density of 32 mg cm⁻³, when empty. In practice, we observe the phantom angled shapes floating in the interphase region between water/methanol above and chloroform below, so the structure that we see in our vials is permeated with a mixture of these solvents.

A related number of interest is the density of vertices. With one (distorted) hexagon per quasi-cell and each of its six vertices shared between three neighboring hexagons in the quasi-plane, there are net two vertices per quasi-cell. This can also be seen from the four rods per quasi-cell, each rod having 1/4 of a vertex at each of its ends, i.e., 1/2 of a vertex net per rod or two vertices per quasi-cell. From the quasi-cell volume in (2) above, we find that the number M of vertices per unit volume is given by

Using our calculated value of h = 4.806 nm, we find M = 5.85 × 10¹⁸ cm⁻³.

Turning to the simple cubic structure, the unit volume of h³ uses only three rods because the 12 rods defining a cube are each shared with four neighboring cubes. For cubes, we would need to devise a sixfold chemical vertex with orthogonal axes. Comparing the efficiency of the above hexagonal diamond to the simple cubic structure as a creator of volume per unit number of polymer rods, the ratio is $43/3=2.309$ in favor of the former. The simple cubic alternative would be expected to produce copious 90° angles, which are not observed.

With the aid of a wire model, it is possible to clearly see two types of open channels running through the hexagonal diamond 2H structure. The first channel type is exactly hexagonal in cross section and runs perpendicular to the quasi-planes. This first set of channels, parallel to the “hexagonal” axis, is intersected by sixfold channels of a second type aligned perpendicularly to it and distributed around it at 60° angular spacings. The latter channels are not exactly hexagonal in cross section, but rather have the cross section shown in Fig. 10(e) or the mirror image of Fig. 10(e). They alternate at 60° angles, yielding an overall threefold rotational symmetry. Altogether, there are seven directions of wide-open channels within the structure that can be fast conduits for molecular diffusion. Other directions would necessitate a zigzag motion and have much slower diffusion rates. The hexagonal direction channels have a slightly larger cross section than the sixfold perpendicular channels in the ratio $3/2=1.225$⁠. The faster degree of diffusion in the hexagonal axis direction may explain the generally linear trend of the structures. Elsewhere, the 60° angles between transverse channels can explain the angled bends that are observed.

With this type of structure, it is possible to understand how a concentration gradient would lead to the delivery of subunit rods in one or a few of the seven internal directions, leading to a surface growth of the structure that was not uniformly directed. Moreover, the angles between growth directions would often differ by 60°, as observed.

Meteoritic polymers of glycine with iron based on a 1494 Da core unit have been assigned rod-like character based upon analysis by mass spectrometry³ and have been noted to assume a variety of characteristic macroscopic morphologies that beg for an explanation. Re-hydration of an extract from Acfer 086 gave x-ray diffracting crystals with a pattern characteristic of a multiple-walled nanotube. In the present nanotube crystal, x-ray diffraction shows a square tiling with a lattice side of 4.836 nm. The (MMFF, 298 K) calculated length of the 1494 Da rod identified in mass spectrometry, between tetraskelion vertices, is 4.838 nm, in very good agreement with the diffraction measurement, also at 295–298 K. The 45° lobed diffraction rings at 6.31 and 5.29 Å are believed to be generated by the square pattern of iron atoms in each tetraskelion vertex. The deep stack of vertices generates a small ring thickness, and the limited lateral iron spacing generates the lobes. However, the calculated pattern is at 5.87 Å, which may indicate that the iron location is split to generate the two rings on either side of this. The nanotube crystals do not necessarily represent structures pre-existing within the meteorite, but are the product of extraction, evaporation, sparing re-hydration, and crystallization.

In a second structural arena, Folch solvent extracts of micrometer-scale meteoritic particles display floating small “phantom-like” tubules with frequent angled bends that suggest the influence of an underlying smaller-scale molecular structure.³ In the same interphase layer where they are observed, there is the highest mass spectrometry signal of glycine/iron polymers of mass around 1600 Da with linear form.³ Here, we have proposed an explanation for the tubules in terms of a space-filling lattice of 1638 Da polymers linked by silicon atoms. We find that the simplest and most efficient space-filling lattice is the diamond 2H structure, with tetrahedral symmetry at the connecting vertices. By analogy with meteoritic triskelia³ and because they have the requisite tetragonal symmetry, we propose that the connecting vertices are silicon atoms. Here, the lattice is permeated by solvent and floating in the layer that matches its density. The structure in free space, without solvents, has a calculated density of 32 mg cm⁻³ and its regular internal passages have angle relationships of 90° and 60° that may guide molecular diffusion in the Folch solution to form the observed macroscopic angles. With such a low-density microscopic structure, we can explain the apparently large volume of tubules relative to the meteoritic particles that released them.

It is possible that the accretion of planetary bodies in the proto-solar molecular cloud could initially have been driven by the development of such a space filling structure in near vacuum out of pre-existing glycine, iron, and silicon.¹ Water, once trapped in such structures, would hydrogen bond preferentially around the more polar vertices and eventually fill the structures, in the cool (100 K) molecular cloud. With slight variations, the polymers can make linear connecting tubes, surround air bubbles with vesicle walls, and fill space with a low-density lattice.

See the supplementary material Sec. S1, which contains the following: (a) description of the sources for the sample material, (b) method for the generation of micron scale meteorite particles (meteorite etch), (c) outline of the research rationale, (d) solvent extraction methods, and (e) crystal mounting and shipping for analysis on the APS synchrotron.

The late Guido Guidotti of Harvard gave encouragement and advice for all of this extraterrestrial polymer research. The Acfer-086 meteorite sample was made available by Raquel Alonso Perez, curator of the Harvard Mineralogical and Geological museum. The KABA meteorite sample was made available at Harvard to J.E.M.M. by the director, Professor, Dr. Béla Baráth, and the deputy director, Dr. Teofil Kovács, of the Museum of the Debreceni Reformatus Kollegium, Kalvin ter 16, H4026, Debrecen, Hungary. We thank Charles H. Langmuir and Zhongxing Chen for the use of the Hoffman clean room in the Harvard Earth and Planetary Science department. Shao-Liang Zheng is thanked for his initial x-ray analyses at the Harvard Department of Chemistry and Chemical Biology. Rachelle Gaudet and Jose Vellila of Harvard are also thanked for advice and assistance on crystallography techniques. The significant x-ray diffraction images were obtained on the APS synchrotron (Argonne, IL) by Anton Joseph Frommelt at APS Argonne. This research used resources of the Advanced Photon Source, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. Use of the Lilly Research Laboratories Collaborative Access Team (LRL-CAT) beamline at Sector 31 of the Advanced Photon Source was provided by Eli Lilly and Company, which operates the facility.

The data that support the findings of this study are available from the corresponding author upon reasonable request.