《自然》（20220616出版）一周论文导读

天文学Astronomy

▲ 作者：Andrew B. Newman, Gwen C. Rudie, Guillermo A. Blanc, Mahdi Qezlou, Simeon Bird, Daniel D. Kelson, et al.

▲ 链接：

https://www.nature.com/articles/s41586-022-04681-6

▲ 摘要：

星系原星团最终将成长为人们在本宇宙中看到的大质量星团，一般通过定位星系的超密度来追踪。对遥远星系的大型光谱调查现已存在，但其灵敏度主要取决于星系的恒星形成活动和尘埃含量，而非它的质量。因此，人们需要不依赖星系成分的大质量原星团的示踪物。

研究组报道了在背景星系密集网格的光谱中观察到莱曼-α吸收，用它通过星系间气体来定位大量红移2.2-2.8的候选原星团。他们发现，与宇宙学模拟中类似物的暗物质含量相比，产生最大吸收的结构（大多数此前未知）包含的星系数量少得惊人。

几乎所有结构都被认为是原星团，研究组推断在该调查中，预期星系成员有一半消失了，因为它们在静止时的紫外线波段异常暗淡。研究组将此归因于原星团环境对这些星系演化的一种出乎意料的强烈早期影响，这些影响减少了它们的恒星形成或增加了它们的尘埃含量。

▲ Abstract：

Galaxy protoclusters, which will eventually grow into the massive clusters we see in the local Universe, are usually traced by locating overdensities of galaxies. Large spectroscopic surveys of distant galaxies now exist, but their sensitivity depends mainly on a galaxy’s star-formation activity and dust content rather than its mass. Tracers of massive protoclusters that do not rely on their galaxy constituents are therefore needed. Here we report observations of Lyman-α absorption in the spectra of a dense grid of background galaxies, which we use to locate a substantial number of candidate protoclusters at redshifts 2.2 to 2.8 through their intergalactic gas. We find that the structures producing the most absorption, most of which were previously unknown, contain surprisingly few galaxies compared with the dark-matter content of their analogues in cosmological simulations. Nearly all of the structures are expected to be protoclusters, and we infer that half of their expected galaxy members are missing from our survey because they are unusually dim at rest-frame ultraviolet wavelengths. We attribute this to an unexpectedly strong and early influence of the protocluster environment on the evolution of these galaxies that reduced their star formation or increased their dust content.

物理学Physics

▲ 作者：Pimonpan Sompet, Sarah Hirthe, Dominik Bourgund, Thomas Chalopin, Julian Bibo, Joannis Koepsell, et al.

▲ 链接：

https://www.nature.com/articles/s41586-022-04688-z

▲ 摘要：

量子多体系统的拓扑结构深刻改变了人们对物质量子相的理解。反铁磁自旋-1霍尔丹链模型在阐明这些效应方面发挥了重要作用。它的基态是一个无序态，由于分数自旋激发，具有对称保护的四重简并边态。

在块体中，霍尔丹链的特征为消失的两点自旋关联、间隙激励和一个特征非局域序参数。最近人们已了解到，霍尔丹链是物质对称保护拓扑相的一种更普遍分类方案，基于与量子信息和纠缠有关的理念。

研究组在一个超冷原子量子模拟器中用费米-哈伯德阶梯实现了这种拓扑霍尔丹相的有限温度版本。通过使用单点和粒子分辨测量以及非局域相关函数，他们直接揭示了系统的边缘和块体特性。

通过不断改变系统的哈伯德相互作用强度，研究组使用一种新型相关器，研究相位对远离海森堡模型的电荷（密度）波动的鲁棒性。

▲ Abstract：

Topology in quantum many-body systems has profoundly changed our understanding of quantum phases of matter. The model that has played an instrumental role in elucidating these effects is the antiferromagnetic spin-1 Haldane chain. Its ground state is a disordered state, with symmetry-protected fourfold-degenerate edge states due to fractional spin excitations. In the bulk, it is characterized by vanishing two-point spin correlations, gapped excitations and a characteristic non-local order parameter. More recently it has been understood that the Haldane chain forms a specific example of a more general classification scheme of symmetry-protected topological phases of matter, which is based on ideas connected to quantum information and entanglement. Here, we realize a finite-temperature version of such a topological Haldane phase with Fermi–Hubbard ladders in an ultracold-atom quantum simulator. We directly reveal both edge and bulk properties of the system through the use of single-site and particle-resolved measurements, as well as non-local correlation functions. Continuously changing the Hubbard interaction strength of the system enables us to investigate the robustness of the phase to charge (density) fluctuations far from the regime of the Heisenberg model, using a novel correlator.

人工智能Artificial Intelligence

▲ 作者：Farshid Ashtiani, Alexander J. Geers & Firooz Aflatouni

▲ 链接：

https://www.nature.com/articles/s41586-022-04714-0

▲ 摘要：

从计算机视觉到医学诊断的深度神经网络通常使用基于时钟的处理器来实现，其中计算速度主要受时钟频率和内存访问时间的限制。在光学领域，尽管光子计算取得了较大进展，但芯片上光学非线性缺乏可扩展以及光子器件的损耗限制了光学深度网络的可扩展性。

研究组报道了一种集成的端到端光子深度神经网络（PDNN），通过直接处理在芯片像素阵列上传播的光波，经神经元层执行亚纳秒级的图像分类。在每个神经元中，线性计算以光学方式执行，非线性激活函数以光电方式实现，允许分类时间低于570?ps，可与最先进数字平台的单时钟周期相媲美。

均匀分布的光源提供了相同的单神经元的光输出范围，允许扩展到大规模PDNN。对手写字母二级分类和四级分类的准确率分别高于93.8%和89.8%。

光学数据的直接、无时钟处理消除了模数转换和对大型内存模块的需求，从而为下一代深度学习系统提供更快、更节能的神经网络。

▲ Abstract：

Deep neural networks with applications from computer vision to medical diagnosis are commonly implemented using clock-based processors, in which computation speed is mainly limited by the clock frequency and the memory access time. In the optical domain, despite advances in photonic computation, the lack of scalable on-chip optical non-linearity and the loss of photonic devices limit the scalability of optical deep networks. Here we report an integrated end-to-end photonic deep neural network (PDNN) that performs sub-nanosecond image classification through direct processing of the optical waves impinging on the on-chip pixel array as they propagate through layers of neurons. In each neuron, linear computation is performed optically and the non-linear activation function is realized opto-electronically, allowing a classification time of under 570?ps, which is comparable with a single clock cycle of state-of-the-art digital platforms. A uniformly distributed supply light provides the same per-neuron optical output range, allowing scalability to large-scale PDNNs. Two-class and four-class classification of handwritten letters with accuracies higher than 93.8% and 89.8%, respectively, is demonstrated. Direct, clock-less processing of optical data eliminates analogue-to-digital conversion and the requirement for a large memory module, allowing faster and more energy efficient neural networks for the next generations of deep learning systems.

材料科学Materials Science

▲ 作者：Hyunjin Kim, Youngjoon Choi, Cyprian Lewandowski, Alex Thomson, Yiran Zhang, Robert Polski, et al.

▲ 链接：

https://www.nature.com/articles/s41586-022-04715-z

▲ 摘要：

魔角扭曲三层石墨烯（MATTG）是一种具有强电子关联和非常规超导性的莫尔材料。然而，对该体系的局部光谱研究仍是不足。

研究组对MATTG进行了高分辨率扫描隧道显微镜和光谱分析，揭示了有利于镜像对称叠加的原子重建的广泛区域。在这些区域观察到对称破缺电子跃迁和掺杂相关的能带结构变形，类似于魔角双层石墨烯，与平带共性理论预期相符。

最值得注意的是，在密度窗口中，每个莫尔单元上有两到三个孔，超导性的光谱特征表现为费米水平的隧穿电导明显下降，伴随着在高温和磁场下逐渐被抑制的相干峰。

观察到的掺杂后电导演化与从间隙超导体到节点超导体的门可调谐跃迁相一致，这在理论上与从巴丁-库珀-施里弗超导体到具有节点序参数的玻色-爱因斯坦凝聚超导体的骤然相变一致。

在这个掺杂窗口内，研究组还检测到峰-谷-隆结构，这表明超导性是由MATTG的玻色子模强耦合驱动的。该结果将有助于人们进一步了解在扭曲双层石墨烯之外，石墨烯基的莫尔结构中的超导性和相关态。

▲ Abstract：

Magic-angle twisted trilayer graphene (MATTG) has emerged as a moiré material that exhibits strong electronic correlations and unconventional superconductivity. However, local spectroscopic studies of this system are still lacking. Here we perform high-resolution scanning tunnelling microscopy and spectroscopy of MATTG that reveal extensive regions of atomic reconstruction favouring mirror-symmetric stacking. In these regions, we observe symmetry-breaking electronic transitions and doping-dependent band-structure deformations similar to those in magic-angle bilayers, as expected theoretically given the commonality of flat bands. Most notably in a density window spanning two to three holes per moiré unit cell, the spectroscopic signatures of superconductivity are manifest as pronounced dips in the tunnelling conductance at the Fermi level accompanied by coherence peaks that become gradually suppressed at elevated temperatures and magnetic fields. The observed evolution of the conductance with doping is consistent with a gate-tunable transition from a gapped superconductor to a nodal superconductor, which is theoretically compatible with a sharp transition from a Bardeen–Cooper–Schrieffer superconductor to a Bose–Einstein-condensation superconductor with a nodal order parameter. Within this doping window, we also detect peak–dip–hump structures that suggest that superconductivity is driven by strong coupling to bosonic modes of MATTG. Our results will enable further understanding of superconductivity and correlated states in graphene-based moiré structures beyond twisted bilayers.

化学Chemistry

▲ 作者：Xin Lu, Zhengyang Zhang, Takehito Hiraki, Osamu Takeda, Hongmin Zhu, Kazuyo Matsubae, et al.

▲ 链接：

https://www.nature.com/articles/s41586-022-04748-4

▲ 摘要：

现今的重熔技术回收铝废料降低了铝的质量，而这种降级回收铝的最终归宿是铝铸造合金。随着更多消费者选择电池驱动的电动汽车而非内燃机汽车，人们对高级铝的需求将大大增加，同时对主要用于内燃机生产的低级再生铝的需求将下降。为了满足未来对高级铝的需求，需要一种新的铝回收方法，能够将废铝提升至与原铝相当的水平。

研究组提出了一种使用熔融盐的固态电解（SSE）工艺来升级回收铝废料。SSE生产的铝纯度可与铝铸造合金中的原铝相媲美。

此外，据估计，工业SSE的能耗不到原铝生产过程的一半。通过有效回收废铝，可以持续满足人们对高级铝的需求。通过使用这种高效、低能耗的工艺，铝循环的真正可持续性指日可待。

▲ Abstract：

The recycling of aluminium scrap today utilizing a remelting technique downgrades the quality of the aluminium, and the final sink of this downgraded recycled aluminium is aluminium casting alloys. The predicted increase in demand for high-grade aluminium as consumers choose battery-powered electric vehicles over internal combustion engine vehicles is expected to be accompanied by a drop in the demand for low-grade recycled aluminium, which is mostly used in the production of internal combustion engines. To meet the demand for high-grade aluminium in the future, a new aluminium recycling method capable of upgrading scrap to a level similar to that of primary aluminium is required. Here we propose a solid-state electrolysis (SSE) process using molten salts for upcycling aluminium scrap. The SSE produces aluminium with a purity comparable to that of primary aluminium from aluminium casting alloys. Moreover, the energy consumption of the industrial SSE is estimated to be less than half that of the primary aluminium production process. By effectively recycling aluminium scrap, it could be possible to consistently meet demand for high-grade aluminium. True sustainability in the aluminium cycle is foreseeable with the use of this efficient, low-energy-consuming process.

地球科学Earth Science

▲ 作者：Sebastian G. Wolf, Ritske S. Huismans, Jean Braun & Xiaoping Yuan

▲ 链接：

https://www.nature.com/articles/s41586-022-04700-6

▲ 摘要：

人们普遍认为碰撞山带地形因地壳增厚而形成，并因河流基岩侵蚀而降低，将气候和构造联系起来。然而，地表过程或岩石圈强度是否控制了山带的高度、形状和寿命仍不确定。此外，如何协调一些活跃造山带的高侵蚀率与山带数亿年的长期生存仍然是个谜。

研究组使用一种新的耦合地表过程和地幔尺度构造的模型来研究山带的生长和衰变。端元模型和新的无量纲Beaumont数（Bm）量化了地表过程和构造如何控制山带的地形演化，并定义了生长造山带的三种端元类型：1型，非稳态、强度控制（Bm>0.5）；2型，通量稳态、强度控制（Bm ≈0.4-0.5）；3型，流量稳态、侵蚀控制（Bm<0.4）。

研究结果表明，喜马拉雅-西藏和安第斯山脉中部（均为1型）的构造占主导地位，台湾岛（可能为2型）的高效地表过程平衡了高收敛速度，而新西兰阿尔卑斯山脉南部（3型）的地表过程则占主导地位。

造山衰变由侵蚀效率决定，可细分为两个阶段，具有可变的均衡回弹特征和相关的时间尺度。该结果提供了一个统一框架，解释了地表过程和岩石圈强度如何控制山带的高度、形状和寿命。

▲ Abstract：

It is widely recognized that collisional mountain belt topography is generated by crustal thickening and lowered by river bedrock erosion, linking climate and tectonics. However, whether surface processes or lithospheric strength control mountain belt height, shape and longevity remains uncertain. Additionally, how to reconcile high erosion rates in some active orogens with long-term survival of mountain belts for hundreds of millions of years remains enigmatic. Here we investigate mountain belt growth and decay using a new coupled surface process and mantle-scale tectonic model. End-member models and the new non-dimensional Beaumont number, Bm, quantify how surface processes and tectonics control the topographic evolution of mountain belts, and enable the definition of three end-member types of growing orogens: type 1, non-steady state, strength controlled (Bm>0.5); type 2, flux steady state, strength controlled (Bm≈0.4-0.5); and type 3, flux steady state, erosion controlled (Bm<0.4). Our results indicate that tectonics dominate in Himalaya–Tibet and the Central Andes (both type 1), efficient surface processes balance high convergence rates in Taiwan (probably type 2) and surface processes dominate in the Southern Alps of New Zealand (type 3). Orogenic decay is determined by erosional efficiency and can be subdivided into two phases with variable isostatic rebound characteristics and associated timescales. The results presented here provide a unified framework explaining how surface processes and lithospheric strength control the height, shape, and longevity of mountain belts.