A research team at Northwestern University arranged about one million microcrystals on a 2 cm square substrate to examine in detail how composition, crystal structure, and light absorption are related in a lead-free semiconductor. At a bromine fraction of around 30%, the band gap, a rough indicator of the energy at which a material begins to absorb light, reached a minimum of 2.11 eV. Inside the crystals, the team found a structure made of a bromine-rich core and an iodine-rich shell. Combining experiments and calculations, the researchers showed how the mixing ratio of elements changes the internal structure of a crystal, and how that in turn affects its optical properties.
The results were published online on November 10, 2026, in the journal Advanced Materials as a peer-reviewed paper by Olivier J. G. L. Chevalier and colleagues.
Arranging one million microcrystals to survey semiconductor compositions
The team studied a family of semiconductors known as "perovskite analogs," which share cesium (Cs) as a common component while varying the ratios of antimony (Sb) and bismuth (Bi), and of bromine (Br) and iodine (I). Their composition formula is Cs₃Sb₂₋ₓBiₓBrᵧI₉₋ᵧ, and they contain no lead. Because the mixing ratios of the metal elements and the halogen elements can be varied independently, an enormous number of compositions is possible. Synthesizing and evaluating samples one by one with conventional methods would require a great deal of time and effort.
The team therefore fabricated a "megalibrary," in which composition changes continuously with position on the substrate. They used a template made of the polymer parylene C, with about one million tiny holes formed at 20 µm intervals over a 2 cm square area. Precursor solutions of different end-member compositions were sprayed from the four corners of the substrate, so that the ratio of bromine to iodine varied in the horizontal direction and the ratio of antimony to bismuth varied in the vertical direction. This allowed crystals spanning a wide range of compositions to form on a single substrate.
The solutions were dried by heating in air at 70°C, and when the parylene C template was peeled away, microcrystals with an average diameter of about 10 µm and a thickness of about 1 µm remained on the substrate. The 70°C figure is the crystal synthesis temperature; the template fabrication process itself involves photolithography, plasma processing, and treatment at 110°C. In total, the researchers made four megalibraries: three with different spraying orders of the precursor solutions, and one for closer study of iodine-rich compositions.
The advantage of packing the microcrystals densely is not only that many compositions can be made at once. In conventional nanocrystal arrays, the amount of sample per square millimeter was only a few nanograms, which was not enough for structural analysis such as X-ray diffraction. With this method, the sample amount could be raised to as much as 0.3 µg/mm², so the team could examine not only light absorption but also elemental composition and crystal structure. By placing the crystals on a flat, transparent substrate, they could link each measurement position to the composition of the material and compare multiple properties on the same composition map.
However, the researchers did not measure all one million crystals individually. According to Section 2.4 of the paper, absorption spectra were measured at 324 points per chip, with each point capturing the average response of about 960 microcrystals. Acquiring one spectrum took about 45 seconds, and the hundreds of resulting measurements were interpolated to compile a map of the relationship between composition and band gap.
X-ray photoelectron spectroscopy (XPS), which probes elemental composition, and X-ray diffraction (XRD), which probes crystal structure, were each measured at 169 points per chip. With this approach, many compositions can be evaluated together, and once compositions with distinctive properties are found, the internal structure of individual crystals can be examined in detail.
A surprising result: more bromine narrows the band gap
Comparing the pure end members, the band gaps were 2.27–2.38 eV for iodine-based materials and 2.59–2.63 eV for bromine-based materials. In common lead halide perovskites, the band gap tends to widen as iodine is replaced with bromine. In the lead-free perovskite analogs examined here, however, the band gap did not change monotonically as the bromine fraction increased.
In the low-bromine region, increasing the bromine fraction first widened the band gap. But once bromine reached roughly 8–14%, depending on the ratio of the metal elements, the gap narrowed sharply, reaching a minimum of 2.11 eV at a composition of about 30%. As the bromine fraction increased further, the band gap widened again. To understand this change, the team investigated the atomic arrangement inside the crystals.
In iodine-dominant materials, octahedra of halogen atoms surrounding metal atoms formed pairs, with each pair isolated from the others, creating a zero-dimensional (0D) structure. Bromine-dominant materials, by contrast, have a two-dimensional (2D) layered structure in which the octahedra are connected more extensively. The terms zero-dimensional and two-dimensional here do not describe the shape of the whole crystal but the directions in which the octahedral atomic arrangement is connected.
In intermediate compositions, multiple peaks and shoulders appeared in the XRD measurements, showing that a 0D phase and a 2D phase coexisted. In other words, bromine and iodine did not mix uniformly within a single crystal structure but separated into regions with different atomic connectivity. The team believes this phase separation forms a 2D structure with relatively weak electron confinement, which narrows the band gap.
The minimum of 2.11 eV, however, is a value obtained from the absorption spectra using a Tauc plot. In samples where multiple phases coexist, the absorption of each phase could not be completely separated. The authors explain that the measured band gap most likely reflects strongly the absorption of the coexisting phase with the narrowest band gap.
To support the experimental results, the team also performed calculations using density functional theory (DFT). Calculations of the free energy of mixing showed that a state in which the 0D and 2D phases coexist is energetically favorable, and that in antimony-rich materials in particular, the energetic penalty for mixing the halogens uniformly was larger than in bismuth-based ones. The authors link this difference to contraction of the bonds around antimony and to structural strain. The actual composition map also confirmed a tendency for phase separation to occur over a wider composition range as the antimony fraction rises.
For this family of materials, a previous study published in Nature Communications in 2025 had also reported non-monotonic changes in band gap with composition and signs of phase separation. The figure of about one million in that study refers to the number of band gap data points obtained from spatially resolved optical measurements. In the present study, by contrast, the team actually fabricated about one million microcrystals and, by combining elemental distribution and crystal structure analysis, investigated even where the different phases form inside each crystal.
How do the bromine-rich core and iodine-rich shell form?
To examine the structure produced by phase separation in detail, the team separately fabricated a microcrystal array of Cs₃SbBiBr₄.₅I₄.₅, in which the ratios of antimony to bismuth and of bromine to iodine are each 1:1.
SEM-EDS analysis, which combines scanning electron microscopy with energy-dispersive X-ray spectroscopy, showed that bromine and antimony were concentrated at the center of the crystal, while iodine and bismuth were concentrated at the periphery. Examining the atomic arrangement by electron diffraction further showed that the central core corresponded to the 2D structure and the surrounding shell to the 0D structure. In other words, the crystals contained a "core-shell structure" in which a bromine-rich 2D core is wrapped in an iodine-rich 0D shell.
Measuring elemental distribution alone does not reveal what crystal structure each region has. Electron diffraction alone, on the other hand, cannot directly identify the distribution of elements. Combining the two measurements showed that the coexistence of two phases confirmed by XRD corresponds to the core and shell inside the crystal.
The team also observed the crystal growth process over time. For the bismuth-containing bromine end member, the precursor has low solubility, and crystals precipitated in under 100 seconds. For the iodine end member, growth took about 500 seconds. In mixed compositions, nuclei formed rapidly during the first 80 seconds or so, and the periphery then grew slowly over several hundred seconds.
From these differences in growth rate, the team proposes a formation process in which the bromine-rich 2D phase precipitates first to form the core, and the iodine-rich 0D phase then grows around it. The difference in solubility between the bromine-based and iodine-based precursors is thought to determine the order of crystallization.
Measurements of the spatial distribution of emission and of fluorescence lifetime also showed that the core and shell have different optical properties. The authors explain that in the 2D structure, where the octahedra are connected, the confinement of electrons and other carriers is weaker than in the 0D structure, so the core has a narrower band gap than the shell. Such core-shell structures have also been observed in bulk samples made by dropping solution and drying it.
A minimum band gap of 2.11 eV, and the stability needed for practical materials
The composition with the narrowest band gap is not necessarily the best material. The team distinguished compositions in which different crystal phases coexist from compositions made of a single phase, and examined the formulation that gives the smallest band gap in each case.
| Category | Composition | Band gap |
|---|---|---|
| Minimum among compositions with coexisting crystal phases | Cs₃Sb₁.₄Bi₀.₆Br₂.₇I₆.₃ | 2.11 eV |
| Minimum among iodine-rich single-phase materials | Cs₃Sb₀.₄Bi₁.₆Br₀.₇I₈.₃ | 2.15 eV |
| Minimum among bromine-rich single-phase materials | Cs₃Sb₁.₆Bi₀.₄Br₆.₅I₂.₅ | 2.21 eV |
Note: All values were obtained from absorption spectra in the same study. Representative compositions described in Section 2.4 of the paper are compared.
The narrowest value, 2.11 eV, was found in a material in which the 0D and 2D phases coexist. Among single-phase materials, the iodine-rich composition gave the minimum at 2.15 eV, and the bromine-rich single-phase material gave 2.21 eV. The suitable composition therefore depends on whether phase uniformity or a narrow band gap is prioritized.
The team also confirmed differences in stability against moisture. When the microcrystal arrays were kept at 25°C and 70% relative humidity and the band gap was measured every three days, degradation over 10 days began in regions rich in antimony and iodine. Regions with a high bromine fraction, in contrast, were relatively stable. Samples whose surface was covered with polystyrene showed greatly reduced degradation over the same period.
The authors explain that microcrystal arrays have a larger surface area exposed to air than thin films do, which is one reason degradation is accelerated. This 10-day test examined the humidity resistance of the microcrystals; it did not measure the lifetime of practical devices such as solar cells.
Although the study investigated the relationships among composition, crystal structure, light absorption, and stability over a wide range, it did not evaluate the power conversion efficiency of solar cells or the operating lifetime of actual devices. Also, the lead-free label concerns the constituent elements; the safety of antimony-containing materials as a whole needs to be assessed separately.
The team believes the megalibrary could also be applied to the search for materials for solar cells, X-ray detectors, photocatalysts, and more. The resulting data on composition and properties may also be usable for machine-learning-driven materials discovery. Verifying the composition-structure relationships found in microcrystals in thin films and real devices could help in selecting materials that combine good optical properties with stability.
