A framework that melts
Metal–organic frameworks (MOFs) are porous crystals of metal ions held together by organic linkers. Zeolitic imidazolate frameworks (ZIFs) are a family of them in which each Zn2+ ion sits at the centre of a tetrahedron of four nitrogen-bound linkers. In ZIF-62 the linkers are a mix of imidazolate (Im, C3H3N2−) and the larger benzimidazolate (bIm, C7H5N2−).
Unlike most MOFs, ZIF-62 melts before it decomposes, between 684 and 719 K, and the melt freezes into a glass very easily. Zinc has a filled 3d10 shell, so the zinc ions themselves do not glow. Any light must come from the linkers.
Schematic, after Fig. 1a: a flat drawing of a tetrahedral network, not the real unit cell. Melting range from the DSC curves (Fig. 1c).
Melt, press, anneal
The crystal is melted and quenched to a melt-quenched glass (MQG). Its glass transition Tg is 593 K. The glass powder is then hot-pressed at 60 MPa by spark plasma sintering, giving a hot-pressed glass (HPG). Pressing loosens the medium-range connectivity of the network: Tg drops to about 550 K.
Finally the pressed glass is annealed at 673 K, 80 K above the Tg of MQG, for 30 min (HPG673-30). Tg recovers to about 563 K: the distorted network relaxes towards a more stable state. X-ray diffraction shows no Bragg peaks in any of the three glasses, and Raman spectra show the same tetrahedral modes in all samples, so the linkers survive intact.
Flow after Fig. 1b; Tg values from the text (Fig. 1c). "1.13 Tg" in the paper means 673 K relative to the 593 K of MQG.
Ultraviolet from the crystal, visible from the glass
Excited at 250 nm, the ZIF-62 crystal gives a narrow emission peak at 300 nm, from the π–π* transition inside the Im and bIm rings. A calculation of the crystal's electronic structure gives flat bands and a gap of 3.58 eV, with both band edges on the carbon and nitrogen p orbitals of the rings. Excited at 320, 370 or 410 nm, the crystal also gives a very weak, broad band at about 400–600 nm.
Melting, pressing and annealing turn the two around. The narrow UV peak weakens, because the rings lose their periodic arrangement. The broad visible band gets much stronger and shifts to longer wavelengths. The authors assign this band to ligand-to-ligand charge transfer (LLCT): an electron excited on one linker relaxes onto a neighbouring one.
Schematic, after Fig. 2a–d and Fig. S6: peak positions as in the text, but heights and band shapes are illustrative and the two bands need different excitation wavelengths.
From blue to cold white
A colour can be placed on the CIE 1931 chromaticity diagram, a map of every colour the eye can see; pure spectral colours lie on its curved edge and white near the middle. Under 370 nm excitation the crystal sits in the blue, at (0.210, 0.216). The melt-quenched glass is even bluer, (0.195, 0.179), and the hot-pressed glass stays close by, (0.205, 0.202).
Annealing makes the jump. HPG673-30 lands at (0.256, 0.295), in the cold-white region. Under a 365 nm laser the change is visible by eye, from a blue glow to a cold-white one. Excited at 250 nm, all samples are much weaker, so 370 nm is the useful excitation.
Coordinates from the text (Fig. 2g). The edge of the diagram uses the tabulated CIE 1931 values; the black-body line is computed here from an analytic fit to the colour-matching functions.
Brighter where it matters, at 365 nm
The internal photoluminescence quantum yield (PLQY) counts the photons emitted per photon absorbed. It depends on the excitation wavelength. At 280 nm, where the UV band dominates, it falls from 4.52% in the crystal to 1.75% in HPG673-30. At 320 nm it first rises from 3.32% to 11.74% on melting, then falls to 5.94% after pressing and annealing.
At 365 nm, the wavelength of a UV LED chip, it rises from 8.04% to 12.19%. At 405 nm it stays at about 12% for all four samples. The emission lifetime at 430 nm also grows through the steps. Heated towards 400 K, a typical LED operating temperature, HPG673-30 loses intensity and lifetime only gradually and linearly.
PLQY values from the text (Fig. 2h, Table S3); bars are drawn only where the text gives a number. Temperature trend from Fig. 2i,j, not drawn.
A disordered, then re-tightened network
Solid-state NMR looks at single atoms. The 67Zn lines of all three glasses are broad, as expected for disordered tetrahedra, and identical to each other, so pressing and annealing do not change the chemistry. The 15N line of Im moves from about 204 to 205 ppm on melting, a sign that σ electrons shift from the linker towards zinc. The imidazole carbons shift too, pointing to charge delocalised over the linkers.
High-energy X-ray scattering gives the distances between atom pairs. The Zn–Zn correlation at 5.94 Å is much weaker in the glasses than in the crystal. But annealing strengthens the short-range ones, C–C(N) at 1.36 Å, Zn–N at 2.02 and 4.17 Å and Zn–C at 3.03 Å, and the first sharp diffraction peak grows: the bonds broken during melting and pressing heal and the medium-range order increases.
Distances from Fig. 4e and the text; the peak heights and the "stronger after annealing" bars are schematic, after Fig. 4e,g.
More, and lower, places for the excitation to land
The authors' explanation runs as follows. In the crystal an electron excited from S0 to S1 returns directly and emits at 300 nm. In the glass, distortion at short and medium range splits the excited level into several lower-lying sublevels, and excitations relax into them before they emit, so the emission moves to the red.
Annealing re-forms Zn–N bonds to the "pyrrole-like" nitrogen of the linkers and pulls the linkers closer. That extends the π conjugation over the tetrahedral unit and opens more LLCT channels, with lower-energy states between the frontier orbitals. The emission position matches the absorption edge of each sample, the red-edge effect.
Schematic, after Fig. 2e and Fig. 5f. The number of dots stands for the population of each level, as in the paper's drawing. The level spacings are illustrative.
The red shift switches on above a threshold
Annealing for 30 min at different temperatures Ta shows a threshold. In the melt-quenched glass the emission peak moves only slightly until Ta = 633 K (1.07 Tg), then shifts sharply. In the pressed glass the jump starts 20 K lower, at 613 K (1.03 Tg): pressing distorts the network more and lowers the barrier for it to relax. The abstract quotes a sharp red shift of 37 nm.
At 633 K the melt has a viscosity of about 108.5 Pa s, extrapolated from literature data with the MYEGA model. That is fluid enough for linkers to rotate and move into better positions within minutes. At 673 K the red shift is complete after about 10 min and stays the same up to 60 min.
Schematic, after Fig. 5d,e: the curves show the shape described in the text, not measured peak positions. Thresholds, viscosity and times from the text.
One glass, one UV chip
Placed on a 365 nm chip at 60 mA, pressed glass (LED-1) gives bluish light with a correlated colour temperature (CCT) of 48 180 K and 1.91 lm/W. The annealed glass (LED-2) gives cold white, 12 382 K and 4.22 lm/W, 2.2 times more. Both render colours better (Ra 85.2 and 81.8) than a commercial white LED with a YAG:Ce phosphor (Ra 68). Adding a commercial red phosphor (LED-3) brings the CCT down to 8 448 K and Ra up to 91.8, at 3.15 lm/W.
The efficacy is still far below the 110 lm/W of the commercial LED. Under continuous driving LED-2 drops from 4.2 to 3.04 lm/W after 3 h and to 1.54 lm/W (36.8%) after 72 h. Heating the glass to 423 K for up to 9 h in air does not change its infrared spectrum, so heat is not the cause; whether UV light creates colour centres is still open.
Values from Table 1 and the text (Fig. 3e). Lines between the three stability points only guide the eye.