Overprediction is a major limitation of current crystal structure prediction (CSP) methods. It is difficult to determine whether computer-predicted polymorphic structures are artefacts of the calculation model or are polymorphs that have not yet been found. Here, we reported the well-known vitamin nicotinamide (NIC) to be a highly polymorphic compound with nine solved single-crystal structures determined by performing melt crystallization. A CSP calculation successfully identifies all six Z′ = 1 and 2 experimental structures, five of which defy 66 years of attempts at being explored using solution crystallization. Our study demonstrates that when combined with our strategy for cultivating single crystals from melt microdroplets, melt crystallization has turned out to be an efficient tool for exploring polymorphic landscapes to better understand polymorphic crystallization and to more effectively test the accuracy of theoretical predictions, especially in regions inaccessible by solution crystallization.

过度预测是当前晶体结构预测 (CSP) 方法的主要限制。很难确定计算机预测的多态结构是计算模型的假象还是尚未发现的多态。在这里，我们报道了众所周知的维生素烟酰胺 (NIC) 是一种高度多晶型化合物，具有九个通过执行熔融结晶确定的已解析单晶结构。CSP 计算成功识别所有六个Z′ = 1 和 2 个实验结构，其中五个无视 66 年来使用溶液结晶进行探索的尝试。我们的研究表明，当结合我们从熔体微滴中培养单晶的策略时，熔体结晶已被证明是探索多晶型景观以更好地理解多晶型结晶并更有效地测试理论预测准确性的有效工具，尤其是在区域中无法通过溶液结晶获得。