Autonomous Off-Equilibrium Morphing Pathways of a Supramolecular Shape-Memory Polymer
超分子形状记忆聚合物的自主非平衡变形途径

The morphing nature of living organisms has served as a constant inspiration for designing synthetic shape-shifting materials with potential impact on a wide range of engineering applications. Despite the progress in the ever-evolving morphing capabilities of synthetic materials, they have fallen short of their biological counterparts in terms of the diversity of the morphing behaviors. For instance, the shape-shifting of cephalopods and the growth of cucumber vines can follow unlimited pathways. In contrast, the pathway of typical synthetic morphing materials is often quite limited. For instance, common responsive hydrogels can only shift between a volume-expanded state and a contracted state in a monotonic way. Similarly, the most basic dual-shape-memory polymer
生物体的变形特性一直是设计合成变形材料 的灵感源泉，对广泛的工程应用具有潜在影响。 尽管合成材料的变形能力在不断进步，但就变形行为的多样性而言，它们仍不及生物材料。例如，头足类动物的变形和黄瓜藤蔓的生长 ，可以遵循无限的路径。与此相反，典型合成变形材料的变形途径往往相当有限。例如，常见的反应性水凝胶只能以单调的方式在体积膨胀状态和收缩状态之间转换。同样，最基本的双形态记忆聚合物

The ORCID identification number(s) for the author(s) of this article can be found under https://doi.org/10.1002/adma.202102473.
本文作者的 ORCID 识别码可在 https://doi.org/10.1002/adma.202102473 下找到。

DOI: 10.1002/adma. 202102473
(SMP) can only switch monotonically between its temporary and permanent shapes. Advanced hydrogels incorporating distinct hydrogel components in a composite manner allow access to much more sophisticated shape-shifting behaviors such as reversible buckling. However, the behavior typically requires complex external triggering, for example, two/ more different stimuli. Similarly, multiSMP capable of complex shifting amongst more than two shapes demands access to multi-step temperature control in both the programming and recovery processes. From the standpoint of device applications, shape programming step is typically conducted outside the device deployment/utilization process. Complex programming thus does not compromise the device potential. On the contrary, the necessary conditions to trigger the device deployment may place severe limitation on the practical values. In other words, a mechanism that allows triggering complex shape-shifting with an easily accessible stimulation method is highly desirable. The recently emerged temporal programming comes into sight. It utilizes time as a naturally available stimulus to control the shape-shifting behavior instead of other conventional physical stimuli such as temperature and . Although elegant, the concept suffers from two notable drawbacks. First, its shapeshifting starts immediately after the programming without a controlling mechanism for its onset. Second, accessing complex autonomous shape-shifting behaviors with temporal programming requires imposing correspondingly sophisticated programming forces. Both these limitations arise from the fact that current temporal shape-shifting systems utilize time as the sole programming parameter to control the shape-shifting behavior.
(SMP) 只能在临时形状和永久形状之间单调切换。先进的水凝胶 以复合的方式将不同的水凝胶成分结合在一起，可以实现更复杂的形状变换行为，如可逆屈曲。不过，这种行为通常需要复杂的外部触发，例如两种或多种不同的刺激。同样，能够在两种以上形状之间进行复杂变换的 multiSMP 需要在编程和恢复过程中进行多步温度控制。 从设备应用的角度来看，形状编程步骤通常在设备部署/使用过程之外进行。因此，复杂的编程不会影响设备的潜力。相反，触发设备部署的必要条件可能会严重限制设备的实用价值。换句话说，我们非常希望有一种机制，能够通过一种容易获得的刺激方法触发复杂的形状变换。最近出现的时间编程便映入眼帘。 它利用时间作为一种自然可用的刺激来控制变形行为，而不是其他传统的物理刺激，如温度和 。这一概念虽然优雅，但有两个明显的缺点。首先，它的变形行为在编程后立即开始，而没有控制其开始的机制。其次，利用时间编程实现复杂的自主变形行为需要施加相应的复杂编程力。目前的时空移形系统利用时间作为唯一的编程参数来控制移形行为，因此产生了上述两个局限性。

With the realization of the universal time-temperature superposition for polymers, we hypothesize that controlling the programming temperature can greatly extend the scope of autonomous shape-shifting. This is non-trivial since temperature control can be achieved digitally in a non-contact spatioselective way via for instance a photothermal mechanism. We also conjecture that a glass transition below room temperature may allow switching on-off the autonomous shapeshifting. With these thoughts in mind, we design a -based network with strong time-temperature dependent ureidopyrimidinone hydrogen bonds (UPy) moieties. Herein, UPy motif is chosen because of its strong quadruple hydrogen
随着聚合物普遍时间-温度叠加的实现，我们假设控制编程温度可以极大地扩展自主变形的范围。这并非难事，因为温度控制可以通过光热机制等非接触式空间选择方式以数字方式实现。 我们还推测，室温以下的玻璃转变可能允许开关自主变形。基于这些想法，我们设计了一种基于 的网络，该网络具有强时间-温度依赖性脲基嘧啶酮氢键（UPy）分子。 之所以选择 UPy 分子，是因为它具有很强的四重氢键。
bonding, the ease of incorporation into a network, and its thermal dissociation at a convenient temperature range. We show that simultaneous control of time and temperature in the programming step leads to unusual freedom in diversifying shape-shifting pathways in a non-monotonic manner. Importantly, the shape-shifting, despite its complexity, proceeds naturally with time (i.e., autonomous), that is, requires no external physical stimulation. The above advantages are reflected in several unusual device designs, including "invisible"-color-based clock, a time-temperature indicator (TTI), and self-evolving sequence-programmable 4D printing.
我们的研究表明，在编程步骤中同时控制时间和温度，能以非单调的方式实现不同寻常的自由形状变换途径。 我们的研究表明，在编程步骤中同时控制时间和温度，能以非单调的方式实现不同寻常的自由形状变换途径。重要的是，尽管形状变换非常复杂，但它随着时间的推移而自然进行（即自主进行），也就是说，不需要外部物理刺激。上述优势体现在几种不同寻常的设备设计中，包括基于颜色的 "隐形 "时钟、时间-温度指示器（TTI）和自演进序列可编程 4D 打印。

The polymer network exhibiting the targeted behavior was synthesized via the base-catalyzed thiol-ene reaction (Figure 1a) between a tetrathiol (PTME; pentaerythritol tetrakis (3-mercaptopropionate)), a UPy acrylate (UPyA), and a diacrylate (BADA; bisphenol A ethoxylate diacrylate). The resulting network contained chemical crosslinks from the thiol-acrylate reaction and physical hydrogen bonding crosslinks from the UPy units (Figure 1b). The NMR and infrared spectroscopy analysis in Figure S1, Supporting Information, verified the design of the polymer structure. We chose a high content of UPyA
通过碱催化的硫醇-烯反应 （图 1a），在四硫醇（PTME；季戊四醇四巯基（3-巯基丙酸酯））、UPy 丙烯酸酯（UPyA）和二丙烯酸酯（BADA；双酚 A 乙氧基二丙烯酸酯）之间合成了具有目标行为的聚合物网络。生成的网络包含硫醇-丙烯酸酯反应产生的化学交联和 UPy 单元产生的物理氢键交联（图 1b）。 NMR 和红外光谱分析（见图 S1，佐证资料）验证了聚合物结构的设计。我们选择了高含量的 UPyA

PTME

UPyA

BADA b

Chemical crosslink
化学交联

Physical crosslink 物理交联

Dissociate 分离

Associate 协理

C

Figure 1. The chemical design, programming principle, and thermomechanical properties of the polymer network. a) Molecular structures of the monomers used in the synthesis. b) The chemical crosslinks and hydrogen bonding physical crosslinks in the network. c) Mechanistic illustration of the programming principle. d) DMA curve of the polymer after eliminating the thermal history at . e) Stress-strain curves obtained at different strain rates at .
图 1.a) 合成所用单体的分子结构。b) 网络中的化学交联和氢键物理交联。c) 编程原理的机理说明。d) 消除 的热历史后聚合物的 DMA 曲线。e) 不同应变速率下的应力-应变曲线。

(40 molar percent in the total acrylates) in order to maximize the time-temperature dependence via the hydrogen bonding, without endangering the covalent integrity of the network. The chemical and physical crosslinking density calculated from the network formulation (see Experimental Section) were 0.69, , respectively.
(占丙烯酸酯总量的 40 摩尔%），以便在不危及网络共价完整性的情况下通过氢键最大限度地提高时间-温度依赖性。根据网络配方计算出的化学和物理交联密度（见实验部分）分别为 0.69 和 。

We aim to explore the strong time-temperature dependence from the hydrogen bonding to establish a unique mechanism for versatile control of the morphing pathways. Before describing that, we note that the material exhibits a classical shape-memory behavior (Figure S2, Supporting Information). When it is stretched at , the strain can be fixed by cooling to . As the sample is returned to , complete strain recovery occurs. Importantly, because the UPy hydrogen bonding is dynamic at ambient temperature, shape recovery is also feasible at albeit at a much slower pace. Unless otherwise noted, we use an identical recovery temperature of throughout this study, as this would allow autonomous recovery (i.e., without extra stimulation) at ambient temperature. This is notably different from common practices that require manipulating the recovery conditions (e.g., temperature) to access versatile shape shifting behaviors. Nevertheless, what deviates the current system from the classical shape-memory is illustrated in Figure 1c. When an external force (e.g., stretching) is applied, all the chain segments are stretched similarly initially but would follow different degrees of hydrogen bonding exchange and chain relaxation depending on the deformation temperature. At a relatively low deformation temperature ( ), macroscopic deformation results in a high percentage of chains being stretched since chain relaxation via the UPy dynamic exchange is limited. This corresponds to a high entropic change, thus the driving force for recovery is large, leading to fast strain recovery (i.e., short recovery time). On the contrary, the UPy dynamic exchange is fast at a higher deformation temperature ( ). This results in partial chain relaxation via the UPy exchange, leading to a lower entropic change, consequently slower recovery and longer recovery time. Because of their different entropic states, the same material can exhibit drastically different shape recovery kinetics, depending on the deformation temperature. Based on this principle, highly versatile non-monotonic shape recovery pathways can be achieved using a homogeneous polymer. We emphasize that the mechanism outlined in Figure 1c is different from creep deformation of a common SMP, which refers to strain increase with time under a constant stress. Often, the entropic change would increase in such a process. For the time-dependent deformation in this work, the strain is kept constant, but the entropic change is reduced via the UPy exchange.
我们的目标是探索氢键对时间-温度的强烈依赖性，从而建立一种独特的机制，实现对变形途径的多功能控制。在说明这一点之前，我们注意到该材料表现出经典的形状记忆行为（图 S2，佐证资料）。当材料在 下拉伸时，应变可通过冷却至 而固定下来。当样品回到 时，应变会完全恢复。重要的是，由于 UPy 氢键在环境温度下是动态的，因此在 下也可以进行形状恢复，尽管速度要慢得多。除非另有说明，否则我们在本研究中始终使用 这一相同的恢复温度，因为这样可以在环境温度下实现自主恢复（即无需额外刺激）。这与需要操纵恢复条件（如温度）才能获得多变的形状变换行为的常见做法明显不同。不过，图 1c 显示了当前系统与经典形状记忆的不同之处。当施加外力（如拉伸）时，所有链段最初都会受到类似的拉伸，但根据变形温度的不同，氢键交换和链松弛的程度也不同。在相对较低的变形温度下（ ），由于通过 UPy 动态交换进行的链松弛受到限制，因此宏观变形会导致高比例的链被拉伸。这相当于高熵变，因此恢复的驱动力大，导致应变恢复快（即恢复时间短）。相反，在较高的变形温度下，UPy 的动态交换速度很快（ ）。这导致通过 UPy 交换产生部分链松弛，从而导致较低的熵变，因此恢复速度较慢，恢复时间较长。由于熵态不同，同一种材料会因变形温度的不同而表现出截然不同的形状恢复动力学。根据这一原理，使用均质聚合物可以实现高度通用的非单调形状恢复途径。我们要强调的是，图 1c 中概述的机制不同于普通 SMP 的蠕变变形，后者是指在恒定应力下应变随时间增加。在此过程中，熵变通常会增加。而本研究中随时间变化的变形过程中，应变保持不变，但熵变却通过 UPy 交换而减少。

DMA analysis shows that the polymer network (Figure 1d) has a glass transition temperature around and a UPy dissociation peak between 40 and . By comparison, the DSC curve (Figure S3, Supporting Information) only shows a at around , corresponding to the overall chain mobility of the network. This observation is in sharp contrast to Meijer's work that shows a melting peak corresponding to UPy stacking for a non-chemically crosslinked polymer, which was absent for our UPy network. This discrepancy is due to the fact that crystallization through UPy stacking requires high molecular chain mobility, which is restricted in our chemically crosslinked network, in consistence with a previous literature report. Based on the above discussion, the second peak in DMA is attributed to the dissociation of hydrogen bonds from UPy, unrelated to UPy crystallization. Tensile testing shows that its mechanical response is highly dependent on the strain rate (Figure 1e) and temperature (Figure S4, Supporting Information). On the contrary, the strain rate dependence of the polymer without UPy is limited (Figure S5, Supporting Information), verifying the strong time temperature dependence from the UPy hydrogen bonding in our network. This establishes the basis for the behaviors outlined in Figure 1c.
DMA 分析表明，聚合物网络（图 1d）的玻璃化转变温度 ，约为 ，UPy 解离峰在 40 到 之间。相比之下，DSC 曲线（图 S3，佐证资料）仅在 左右显示出 ，与网络的整体链流动性相对应。这一观察结果与 Meijer 的研究 形成了鲜明对比，后者的研究显示，非化学交联聚合物会出现与 UPy 堆积相对应的熔化峰，而我们的 UPy 网络却没有这种现象。造成这种差异的原因是，通过 UPy 堆叠结晶需要较高的分子链流动性，而在我们的化学交联网络中，分子链流动性受到限制，这与之前的文献报告一致。 根据上述讨论，DMA 中的第二个峰是由于 UPy 的氢键解离造成的，与 UPy 结晶无关。拉伸测试表明，其机械响应高度依赖于应变速率（图 1e）和温度（图 S4，佐证资料）。相反，不含 UPy 的聚合物对应变速率的依赖性有限（图 S5，佐证资料），这验证了我们的网络中 UPy 氢键对时间温度的强烈依赖性。这为图 1c 中概述的行为奠定了基础。

Figure 2a (corresponding full shape-memory cycles in Figure S6, Supporting Information) shows that, at the same recovery temperature of the strain recovery kinetics of the polymer deformed at different temperatures is indeed quite different. At an identical recovery time of , the recovery ratio decreases drastically with the increase of deformation temperature until reaching a plateau value of at around (Figure 2b). This high dependence of the recovery behavior on the deformation temperature is visually illustrated via the photographic images shown as the inset in Figure 2b. Given the timetemperature superposition, we anticipate that deforming the polymer at an identical temperature for different durations can also be explored as an alternative strategy to control the recovery kinetics. Indeed, changing the deformation time at can alter the recovery kinetics (Figure S7, Supporting Information). Nevertheless, Figure summarizes the time required for full strain recovery corresponding to different deformation temperatures and deformation time. At an identical deformation temperature of , the three colored bars show that it requires a longer recovery time for a sample held under the same deformation strain for longer (i.e., longer deformation time). This phenomenon is also valid for deformation temperatures of 40 and , although the recovery time range differs.
图 2a（相应的完整形状记忆循环见图 S6，佐证资料）显示，在 的相同恢复温度下，不同温度下变形的聚合物的应变恢复动力学确实存在很大差异。在相同的恢复时间 下，恢复比随着变形温度的升高而急剧下降，直到 左右达到 的高位值（图 2b）。图 2b 插图中的照片直观地说明了恢复行为对变形温度的高度依赖性。考虑到时间-温度叠加，我们预计在相同温度下对聚合物进行不同持续时间的变形也可作为控制恢复动力学的另一种策略。事实上，改变 的变形时间可以改变恢复动力学（图 S7，佐证资料）。不过，图 总结了不同变形温度和变形时间下完全恢复应变所需的时间。在相同的变形温度（ ）下，三条彩色柱状图显示，在相同变形应变下保持更长时间（即变形时间更长）的样品需要更长的恢复时间。这一现象同样适用于变形温度为 40 和 的情况，只是恢复时间的范围有所不同。

The deformation temperature determined shape recovery kinetics offers a unique way to control the autonomous morphing pathway. To achieve this goal, we employ a flexible photothermal method by laser printing, in which the digital ink pattern determines the temperature distribution of a sample when it is exposed to infrared light. Specifically, with 1 min irradiation, the temperature increases linearly with the gray scale of the ink within 32 and (Figure S8, Supporting Information). Figure illustrates the morphing versatility achieved with this photothermal method. In all the demonstrations in Figure 2d, the polymer samples are stretched to strain under the ambient condition for and the deformation is fixed by cooling in a refrigerator . As a baseline reference, the photos on the first row illustrate the recovery pathway of a sample deformed without any other additional interference. The rectangular shape recovers in a monotonic way, as would be expected for a classical equilibrium driven shape-memory behavior. The photos on the second row display the recovery pathway for a sample deformed with additional infrared exposure under an otherwise identical condition. The result of the light exposure during the deformation step is that, as the sample recovers, it rolls up and then returns to the original flat state, deviating drastically from the equilibrium-driven pathway on the first row. This unusual behavior is because the infrared light creates a higher temperature in the ink region,
变形温度决定形状恢复动力学为控制自主变形途径提供了一种独特的方法。为了实现这一目标，我们采用了一种灵活的激光打印光热法，当样品暴露在红外光下时，数字油墨图案决定了样品的温度分布。 具体来说，照射 1 分钟后，温度会随着油墨灰度在 32 和 范围内线性上升（图 S8，佐证资料）。图 展示了这种光热方法实现的变形多样性。在图 2d 的所有演示中，聚合物样品在 的环境条件下被拉伸至 应变，然后在冰箱中冷却以固定变形 。作为基线参考，第一行的照片展示了没有任何其他干扰的变形样品的恢复路径。矩形形状以单调的方式恢复，这是经典平衡驱动形状记忆行为的预期效果。第二行的照片显示了在其他条件完全相同的情况下，经过额外红外线照射的变形样品的恢复路径。在变形过程中进行光照射的结果是，样品在恢复过程中会卷起，然后又恢复到原来的平整状态，与第一行的平衡驱动路径大相径庭。这种不寻常的行为是因为红外光在油墨区域产生了较高的温度、

b

d

e

Recovery 恢复

Figure 2. Designing autonomous morphing pathways by controlling the deformation/programming temperature. a) The dependence of strain recovery kinetics on the deformation temperature. b) The recovery ratio ( at ) obtained for different deformation temperatures. The initial lengths for the samples are . c) The time required to reach full recovery for samples deformed under different conditions (temperature and time). d) The autonomous recovery pathways of an ink-printed polymer sample (the sample is and the ink is in the center). e) Complexed autonomous morphing pathways defined by ink and cut patterns. All scale bars are .
图 2.a) 应变恢复动力学与变形温度的关系。 b) 不同变形温度下的恢复比 ( at ) 。样品的初始长度为 。 c) 在不同条件（温度和时间）下变形的样品达到完全恢复所需的时间。 d) 油墨印刷聚合物样品的自主恢复路径（样品为 ，油墨在中心 ）。 e) 由油墨和切割图案定义的复杂自主变形路径。所有比例尺均为 。

leading to a contrast in the recovery speed between areas with ink and without ink.
导致有墨水和无墨水区域的恢复速度形成鲜明对比。

The photos on the third row show another possibility. The sample is subjected to a deformation process identical to the second row except that it is cut along the edges of the ink pattern parallel to the stretching direction. Here, only the ink region bulges out of plane. In comparing the morphing sequence on these three rows, the starting and ending shapes are all planar rectangular shapes, yet their morphing pathways are quite different. This is particularly notable as the polymer used is an identical homogeneous material and the recovery condition is also identical, that is, autonomous without any extra stimulation. Besides the fully autonomous recovery, it is possible to utilize the ink for additional control by employing light as a stimulation. This is shown as the photos on the fourth row. The sample treatment is identical to the third row, except that infrared light is irradiated onto the sample at mark during its recovery process. As a result of this interference, the recovery route is reversed. The ink region, because of its higher recovery temperature, recovers faster than the non-ink regions. Quantitative strain recovery kinetics at 25 and presented in Figure S9, Supporting Information confirms that a higher recovery temperature does lead to faster recovery. When the light is removed after , the sample returns to its original morphing path.
第三行的照片显示了另一种可能性。除了沿着与拉伸方向平行的油墨图案边缘切割之外，样品的变形过程与第二行相同。在这里，只有油墨区域凸出平面。比较这三行的变形顺序，起点和终点形状都是平面矩形，但它们的变形路径却截然不同。这一点尤其值得注意，因为所使用的聚合物是完全相同的均质材料，而且复原条件也完全相同，即无需任何额外刺激即可自主复原。除了完全自主的复原外，还可以利用光作为刺激，对墨水进行额外的控制。如第四行的照片所示。样品处理过程与第三行相同，只是在 刻度处的样品恢复过程中照射了红外线。在这种干扰下，复原路线发生了逆转。由于油墨区域的恢复温度较高，因此其恢复速度快于非油墨区域。图 S9（佐证资料）中 25 和 下的定量应变恢复动力学证实，恢复温度越高，恢复速度越快。当 后移除光源时，样品恢复到原来的变形路径。

Based on the base behaviors established in Figure 2d, more sophisticated morphing behaviors can be designed by the combination of cutting and variation in the ink patterns. Figure displays that an eight-petal flower closes and then reopens in a specific sequence. Again, these sequential actions are rather notable given that the morphing is fully autonomous. Similarly, an arch bridge bugles up in the center and then bugles on the edges upon irradiated by the light, eventually returns to its planar geometry. What is also noteworthy is that despite the out of place complex morphing in Figure 2d,e, the initial programming forces are all much simpler in-plane stretching. The ink patterns and stretching methods corresponding to Figure are shown in Figure S10, Supporting Information.
在图 2d 所建立的基本行为基础上，可以通过切割和墨迹图案变化的组合设计出更复杂的变形行为。图 显示，一朵八瓣花按特定顺序闭合，然后重新开放。同样，由于变形是完全自主的，这些顺序动作也相当引人注目。同样，一座拱桥在光线照射下，先是在中间发出 "嗡嗡 "声，然后在边缘发出 "嗡嗡 "声，最后又恢复到平面几何形状。同样值得注意的是，尽管图 2d、e 中的变形非常复杂，但最初的编程力都是简单得多的平面内拉伸。与图 相对应的油墨图案和拉伸方法见图 S10，《佐证资料》。

b

C

f

Fresh, Fresh 新鲜，新鲜

Bad, Fresh 坏的，新鲜的

Bad, Bad 糟糕，糟糕

Figure 3. Time evolution of stress and colors. a) Isostrain stress relaxation curves at different temperatures (strain: 50%). b) Time evolution of birefringent color and its use to design an "invisible" clock. c) Birefringent pattern evolution at . d) Birefringent pattern evolution at . e) The autonomous evolution of structural colors. f) A time-temperature indicator for monitoring the thermal history of milk. All scale bars are .
图 3：应力和颜色的时间演变。a) 不同温度下的等应变应力松弛曲线（应变：50%）。 b) 双折射颜色的时间演变及其用于设计 "隐形 "时钟。 c) 的双折射图案演变。 d) 的双折射图案演变。 e) 结构颜色的自主演变。 f) 用于监测牛奶热历史的时间-温度指示器。所有刻度条均为 。

So far, we have focused on the contrast in the recovery speed arisen from the difference in the deformation temperature. Besides the benefit of controlling the morphing pathway, we realize that the same underlying mechanism outlined in Figure 1c suggests another unique possibility to control the stress, that is, a higher deformation temperature corresponds to faster hydrogen bond exchange, consequently faster stress relaxation. Indeed, Figure 3a shows that when the polymer is subjected to isostrain stress relaxation experiments, the stress decay is faster at a higher temperature. Based on these data, the activation energy can be calculated as (see the Arrhenius curve in Figure S11, Supporting Information). This offers an opportunity to construct a time-temperature clock. To do so, the polymer is stretched by and fixed at that strain at . The stretched polymer can show invisible color under polarized light due to the birefringence induced by the stress in the material. As the stress decays due to the hydrogen bonding exchange, the residual stress evolves with time. Because the birefringence corresponds to the stress, the color under polarized light would also change due to the stress relaxation (Figure 3b). We emphasize that, due to the existence of the permanent crosslinks, the stress would not relax completely. Correspondingly, the birefringent color eventually becomes white instead of black (zero stress state) (Figure 3b). A video capturing the color evolution is provided as Movie S1, Supporting Information. Since the polymer remains transparent under ambient natural light, this is essentially an "invisible" clock.
到目前为止，我们主要关注的是变形温度差异所引起的恢复速度对比。除了控制变形途径的益处之外，我们还认识到图 1c 中概述的相同基本机制为控制应力提供了另一种独特的可能性，即变形温度越高，氢键交换越快，应力松弛也就越快。事实上，图 3a 显示，当聚合物进行等应变应力松弛实验时，温度越高，应力衰减越快。根据这些数据，可以计算出活化能为 （见图 S11 中的 Arrhenius 曲线，佐证资料）。这为构建时间-温度时钟提供了机会。为此，可通过 拉伸聚合物，并在 固定该应变。拉伸后的聚合物在偏振光下会呈现出不可见的颜色，这是由于材料中的应力引起了双折射。由于氢键交换导致应力衰减，残余应力随时间变化。由于双折射与应力相对应，偏振光下的颜色也会因应力松弛而发生变化（图 3b）。 我们要强调的是，由于永久交联的存在，应力不会完全松弛。相应地，双折射颜色最终会由黑色变为白色（零应力状态）（图 3b）。颜色演变的视频见 "佐证资料 "中的 "影片 S1"。由于聚合物在周围自然光下保持透明，因此这基本上是一个 "隐形 "时钟。

Additional control of the birefringent color evolution is possible by further exploring the temperature dependence of the stress relaxation. A polymer film was stretched by at . While maintaining this strain, the film was locally heated with a metal stamp for with the contacting force kept minimum to avoid physical deformation. This created a stress pattern in the film due to the different speed of stress relaxation between the two regions, which is invisible under natural
通过进一步探索应力松弛的温度依赖性，还可以对双折射颜色演变进行进一步控制。 拉伸聚合物薄膜， 。在保持这种应变的同时，用 金属印章局部加热薄膜， ，接触力保持最小，以避免物理变形。由于两个区域的应力松弛速度不同，这就在薄膜上形成了应力模式，而这种模式在自然条件下是不可见的。
light. Figure 3c and Movie S2, Supporting Information displays the birefringent pattern evolution at this temperature , at a speed higher than that in Figure 3b due to the faster stress decay. The birefringent pattern evolution can be further accelerated by changing the stress relaxation temperature to (Figure 3d). In Figure 3c,d, the patterns eventually vanished because all the areas reached the same degree of stress relaxation (the maximum degree of relaxation that can be achieved), thus the spatial color contrast was erased. We note that, utilizing the transition, any stress states (and the birefringent patterns) shown above can be locked into the polymer by cooling in a refrigerator .
光。图 3c 和 "佐证资料 "中的影片 S2 显示了双折射图案在该温度下的演变过程 ，由于应力衰减速度更快，双折射图案的演变速度比图 3b 中的更快。将应力松弛温度改为 ，可进一步加速双折射图案的演变（图 3d）。在图 3c、d 中，由于所有区域都达到了相同的应力松弛程度（可达到的最大松弛程度），图案最终消失了，因此空间颜色对比消失了。我们注意到，利用 过渡，任何应力状态（以及上述双折射图案）都可以通过在冰箱中冷却而锁定在聚合物中 。

The sensitivity to both time and temperature opens up an ideal opportunity to construct a TTI. To do so, a nanosized dot array pattern was molded onto the polymer surface, giving rise to structural color. Macro-sized letters of "MILK" were embossed onto the film, with "MI" at and "LK" at for . All four letters could be locked by cooling in a refrigerator. Upon returning the film to ambient conditions, the four letters disappeared with time, notably at different paces depending on the stamping temperatures. Changing the embossing temperature can alter the sequence of the letter disappearance, as is captured in Movie S3, Supporting Information. The TTI device was further used to monitor the thermal history of milk (Figure 3f). Specifically, two bottles of fresh milk were labeled with identical TTI labels. The one on the right was stored in the refrigerator and the other under ambient condition. After , the "tick" symbol on the left disappeared while the one in the refrigerator remains intact. If kept under ambient conditions for another , the symbol on right bottle would also disappear. Clearly, this change in labels reflects the thermal history, which is closely related to the quality of the milk. We note that if there are no dynamic hydrogen bonds in the network, the stress relaxation hardly happens and the overall effect described above is not observed.
对时间和温度的敏感性为构建 TTI 提供了理想的机会。 为此，我们在聚合物表面模塑了纳米级的点阵图案，从而产生了结构色。在薄膜上压印了 "MILK "的巨型字母，其中 "MI "位于 ，"LK "位于 ， 。所有四个字母都可以在冰箱中冷却后锁定。将胶片放回环境中后，这四个字母会随着时间的推移而消失，不同的压印温度会有不同的消失速度。改变压印温度可改变字母消失的顺序，见 "佐证资料 "中的影片 S3。TTI 设备还用于监测牛奶的热历史（图 3f）。具体来说，两瓶鲜奶贴上了相同的 TTI 标签。右边的一瓶保存在冰箱中，另一瓶保存在环境条件下。 后，左边的"√"符号消失，而冰箱中的"√"符号保持不变。如果再在常温条件下保存 ，右边瓶子上的符号也会消失。显然，标签的这种变化反映了热历史，而热历史与牛奶的质量密切相关。我们注意到，如果网络中没有动态氢键，应力松弛几乎不会发生，也就观察不到上述整体效应。

To extend the potential utility further, we next proceed to demonstrate the benefit of our material principle with 3D printed devices. To make the material printable by the Digital Light Processing (DLP) method, the material system needs to be adjusted to enable UV curing. This can be done by employing photoinduced radical polymerization of thiol and acrylate, with only slight tuning of the starting precursors. Specifically, we used formulations similar to the one in Figure 1a except that the value of in the BADA monomer was altered in order to tune the (Figure 4a). The polymer network corresponding to has a closest to the polymer network synthesized by the base catalyzed click reaction. This polymer was chosen for further investigation. Changing its deformation temperature and time allowed adjusting the strain recovery kinetics (Figure 4b), similar to what was presented earlier. Importantly, the versatility of 3D printing allows producing diverse complex shapes (Figure 4c). Since the polymer network has shape-memory characteristics, we anticipate that this can be extended to printing, in which time is the a
为了进一步扩大潜在的实用性，我们接下来将用 3D 打印设备来证明我们的材料原理的益处。为了使这种材料能够通过数字光处理（DLP）方法打印，需要对材料系统进行调整，使其能够进行紫外线固化。这可以通过使用硫醇和丙烯酸酯的光诱导自由基聚合来实现，只需对起始前体稍作调整即可。具体来说，我们使用的配方与图 1a 中的配方类似，只是改变了 BADA 单体中 的值，以调整 （图 4a）。与 相对应的聚合物网络 与碱催化点击反应合成的聚合物网络最为接近。我们选择了这种聚合物进行进一步研究。改变其变形温度和时间可以调整应变恢复动力学（图 4b），这与前面介绍的情况类似。重要的是，3D 打印的多功能性允许生产出各种复杂的形状（图 4c）。由于聚合物网络具有形状记忆特性，我们预计这种特性可以扩展到 打印，其中时间是关键。

b

C

Figure 4. Self-evolving sequence programmable 4D printing. a) Dynamic mechanical analysis curves of the UV cured polymer networks. b) Strain recovery kinetics of a polymer network ( ) corresponding to different deformation times and temperatures. c) 3D printed objects. d) Controlled autonomous morphing pathways of an octopus and hands, localized programming conditions are i: ; ii: ; iii: plus . All scale bars are .
图 4：自演化序列可编程 4D 印刷。a) 紫外线固化聚合物网络的动态力学分析曲线。 b) 不同变形时间和温度下聚合物网络的应变恢复动力学 ( ) 。 c) 3D 打印对象。 d) 章鱼和手的受控自主变形路径，局部编程条件为 i: ； ii: ； iii: 加 。所有比例尺均为 。
fourth dimension. Figure 4d illustrates that, by manipulating the deformation temperature and time, the legs of the printed octopus can be programmed to exhibit sequential autonomous morphing. Similarly, the fingers on the two printed identical hands can be programmed to show very different actions that mimic finger-guessing (Movie S4, Supporting Information). We emphasize that the capabilities demonstrated in Figure 4d extend beyond conventional shape-memory-based 4D printing. The key difference lies in that the printed structures can not only morph with time in an autonomous fashion, but that the sequence of the morphing can be controlled with precision. In addition, the complexed morphing is achieved with one homogeneous material, instead of multimaterials.
四维空间。图 4d 显示，通过控制变形温度和时间，可以对打印章鱼的腿进行编程，使其表现出连续的自主变形。同样，两只打印出来的相同手掌上的手指也可以通过编程表现出截然不同的动作，从而模仿手指猜谜（影片 S4，佐证资料）。我们要强调的是，图 4d 中展示的功能超越了传统的基于形状记忆的 4D 打印。其主要区别在于，打印结构不仅可以随时间自主变形，而且变形顺序可以精确控制。此外，复合变形是通过一种同质材料而不是多种材料实现的。

Typical shape-shifting materials exhibit limited morphing pathways unless complexed multimaterials or external stimulation schemes are involved. In contrast, the current work explores the strong time-temperature dependence of supramolecular polymer networks to access off-equilibrium geometric states as an effective way to design versatile morphing pathways. Although we focus on UPy hydrogen bonding for illustrating the concept, many other supramolecular bonds with similar dissociation behaviors can be potential candidates. Overall, our unique approach allows achieving autonomous morphing with a single homogeneous material, greatly simplifying the material and stimulation requirement while expanding the morphing versatility. At the device level, the benefits of our material principle are illustrated with an "invisible" clock, a TTI, and sequence-programmable 4D printing. These device examples, however, only represent a small fraction of the vast potential that has yet to be uncovered in the future.
典型的变形材料显示出有限的变形途径，除非涉及复杂的多材料或外部刺激方案。与此相反，目前的研究工作探索了超分子聚合物网络对时间-温度的强烈依赖性，以获得非平衡几何状态，作为设计多功能变形途径的有效方法。虽然我们以 UPy 氢键为重点来说明这一概念，但其他许多具有类似解离行为的超分子键 也可能是潜在的候选键。总之，我们的独特方法允许使用单一均质材料实现自主变形，大大简化了对材料和刺激的要求，同时扩大了变形的多样性。在设备层面，我们用 "隐形 "时钟、TTI 和序列可编程 4D 打印来说明我们的材料原理的优势。然而，这些设备示例仅代表了未来有待发掘的巨大潜力中的一小部分。

Materials: PTME and BADA ( expectively) were purchased from Sigma Aldrich. N,N-dimethylformamide (DMF), and acetic acid were acquired from Sinopharm. Triethylamine (TEA) and Irgacure 819 were obtained from TCI. UPy containing acrylate monomer (UPyA) was synthesized according to the literature.
材料PTME 和 BADA ( expectively) 购自 Sigma Aldrich。N,N-二甲基甲酰胺（DMF）和乙酸购自国药集团。三乙胺（TEA）和 Irgacure 819 购自 TCI。含有丙烯酸酯单体的 UPy（UPyA）是根据文献合成的。

Synthesis of the Polymer Network: PTME , UPyA , and TEA were mixed with DMF ( ). The mixture was stirred at until UPyA was completely dissolved. BADA ( ) was subsequently added. This precursor solution was poured into a mold defined by two glass slides separated by a silicon rubber spacer thick) and the reaction proceeded in an oven at for . The obtained film was vacuum-dried overnight. The chemical and physical crosslinking densities were calculated from the network formulation as the mole number of BADA per unit total mass and the mole number of UPy per unit total mass divided by 2 , respectively.
聚合物网络的合成：将 PTME 、UPyA 和 TEA 与 DMF ( ) 混合。在 下搅拌混合物，直到 UPyA 完全溶解。随后加入 BADA ( )。将前驱体溶液倒入由两块玻璃片组成的模具中，玻璃片之间用硅橡胶间隔 ），然后在 的烘箱中进行反应， 。得到的薄膜在 真空干燥过夜。根据网络配方计算出的化学和物理交联密度分别为单位总质量中 BADA 的摩尔数和单位总质量中 UPy 的摩尔数除以 2。

Fabrication of the Structural Colored Film: The monomer precursor solution was poured onto a silicon nanotemplate. It was cured at for . The obtained film was vacuum-dried overnight. The surface microstructures were characterized by a confocal laser scanning microscope (Zeiss LSM700) under the "Z-stack" mode.
制作结构性彩色薄膜：将单体前驱体溶液倒在硅纳米模板上。在 下固化 。获得的薄膜经过真空干燥 过夜。用共聚焦激光扫描显微镜（Zeiss LSM700）在 "Z-stack "模式下对表面微结构进行表征。

3D Printing: PTME ( ), UPyA (5.12 g), and TEA ( ) were mixed with DMF ( ). The mixture was stirred at until UPyA was completely dissolved. Then, acetic acid ( ), BADA ( , , Irgacure , and Sudan III were added in the resulting precursor solution. A custom DLP printer using a commercial projector (DELL ) was used for printing at . All the models were printed using a slice thickness of , the exposure time for each layer was , and black time required for recoating was . The
三维打印：将 PTME ( ) 、UPyA（5.12 克）和三乙醇胺 ( ) 与 DMF ( ) 混合。在 下搅拌混合物，直到 UPyA 完全溶解。然后，在得到的前体溶液中加入醋酸 ( ) 、BADA ( , , Irgacure , 和苏丹 III 。使用商用投影仪（戴尔 ）定制的 DLP 打印机在 进行打印。所有模型的打印切片厚度为 ，每层的曝光时间为 ，重新涂层所需的黑色时间为 。打印

for  对于

Digital Temperature Programming by Photothermal Effect: The ink patterns were designed with CorelDRAW X4 software and printed onto the films using a laser printer (HP Laserjet 200 color MFP M276nw). The films were exposed to flood light emitted by a xenon lamp (CEL-
通过光热效应进行数字温度编程：使用 CorelDRAW X4 软件设计墨水图案，并用激光打印机（HP Laserjet 200 color MFP M276nw）打印到薄膜上。将薄膜置于由氙灯（CEL

temperatures were attained with a Fluke Ti10 thermal imager.
温度是用 Fluke Ti10 热成像仪测得的。

Shape-Memory Measurement: All quantitative shape-memory curves were obtained using a DMA Q800 machine under a "strain rate" mode. For Figure 2c, the full recovery times were evaluated by manual measurement of the sample length change.
形状记忆测量：所有定量形状记忆曲线都是使用 DMA Q800 机器在 "应变率 "模式下获得的。对于图 2c，完全恢复时间是通过手动测量样品长度变化来评估的。

Tensile Tests: The mechanical stretching experiments were carried out using a universal material testing machine (Zwick/Roell Z005).
拉伸试验：机械拉伸实验使用万能材料试验机（Zwick/Roell Z005）进行。

Supporting Information is available from the Wiley Online Library or from the author.
辅助信息可从 Wiley 在线图书馆或作者处获取。

The authors would like to thank the following programs for their financial support: National Natural Science Foundation of China (grant nos. 52033009 and 21625402 .
作者感谢以下项目的资助：国家自然科学基金（批准号：52033009 和 21625402 .

The authors declare no conflict of interest.
作者声明没有利益冲突。

Research data are not shared.
研究数据不共享。

4D printing, shape-memory polymers, supramolecular bonds, timetemperature indicators, ureidopyrimidinone
4D打印、形状记忆聚合物、超分子键、时间温度指示器、脲基嘧啶酮

Received: March 31, 2021
收到：2021 年 3 月 31 日

Revised: May 12, 2021 Published online:
已修订2021 年 5 月 12 日在线发布：

[1] Y. Forterre, J. M. Skotheim, J. Dumais, L. Mahadevan, Nature 2005, .
[1] Y. Forterre, J. M. Skotheim, J. Dumais, L. Mahadevan, Nature 2005, .

[2] T. J. White, D. J. Broer, Nat. Mater. 2015, 14, 1087.
[2] T. J. White, D. J. Broer, Nat.Mater.2015, 14, 1087.

[3] L. lonov, Adv. Funct. Mater. 2013, 23, 4555.
[3] L. lonov, Adv.Funct.Mater.2013, 23, 4555.

[4] A. R. Studart, Angew. Chem., Int. Ed. 2015, 54, 3400.
[4] A. R. Studart, Angew.Chem.Ed.2015, 54, 3400.

[5] A. Lendlein, R. Langer, Science 2002, 296, 1673.

[6] S. Xu, Z. Yan, K.-I. Jang, W. Huang, H. Fu, J. Kim, Z. Wei, M. Flavin, J. McCracken, R. Wang, A. Badea, Y. Liu, D. Xiao, G. Zhou, J. Lee, H. U. Chung, H. Cheng, W. Ren, A. Banks, X. Li, U. Paik, R. G. Nuzzo, Y. Huang, Y. Zhang, J. A. Rogers, Science 2015, 347, 154.
[6] S. Xu、Z. Yan、K. -I.Jang, W. Huang, H. Fu, J. Kim, Z. Wei, M. Flavin, J. McCracken, R. Wang, A. Badea, Y. Liu, D. Xiao, G. Zhou, J. Lee, H. U. Chung, H. Cheng, W. Ren, A. Banks, X. Li, U. Paik, R. G. Nuzzo, Y. Huang, Y. Zhang, J. A. Rogers, Science 2015, 347, 154.

[7] B. Jin, H. Song, R. Jiang, J. Song, Q. Zhao, T. Xie, Sci. Adv. 2018, 4, eaao3865.

[8] M. Zarek, M. Layani, I. Cooperstein, E. Sachyani, D. Cohn, S. Magdassi, Adv. Mater. 2016, 28, 4449.
[8] M. Zarek、M. Layani、I. Cooperstein、E. Sachyani、D. Cohn、S. Magdassi，Adv. Mater.2016, 28, 4449.

[9] S. J. Gerbode, J. R. Puzey, A. G. McCormick, L. Mahadevan, Science
[9] S. J. Gerbode、J. R. Puzey、A. G. McCormick、L. Mahadevan，《科学》。

[10] V. Lutz-B, S. Bolisetty, P. Azzari, S. Handschin, R. Mezzenga, Adv. Mater. 2020, 32, 2004941.
[10] V. Lutz-B、S. Bolisetty、P. Azzari、S. Handschin、R. Mezzenga，Adv. Mater.2020, 32, 2004941.

[11] J. Kim, J. A. Hanna, M. Byun, C. D. Santangelo, R. C. Hayward, Science 2012, 335, 1201

[12] Q. Zhao, X. Yang, C. Ma, D. Chen, H. Bai, T. Li, W. Yang, T. Xie, Mater. Horiz. 2016, 3, 422.
[12] Q. Zhao, X. Yang, C. Ma, D. Chen, H. Bai, T. Li, W. Yang, T. Xie, Mater.Horiz.2016, 3, 422.

[13] L. Huang, R. Jiang, J. Wu, J. Song, H. Bai, B. Li, Q. Zhao, T. Xie, Adv. Mater. 2017, 29, 1605390.
[13] L. Huang, R. Jiang, J. Wu, J. Song, H. Bai, B. Li, Q. Zhao, T. Xie, Adv. Mater.2017, 29, 1605390.

[14] T. Xie, Nature 2010, 464, 267.
[14] T. Xie，《自然》，2010，464，267。

[15] T. Ware, K. Hearon, A. Lonnecker, K. L. Wooley, D. J. Maitland, W. Voit, Macromolecules 2012, 45, 1062.

[16] X. Hu, J. Zhou, M. Vatankhah-V, W. F. M. Daniel, Q. Li, A. P. Zhushma, A. V. Dobrynin, S. S. Sheiko, Nat. Commun. 2016, 7, 12919.
[16] X. Hu、J. Zhou、M. Vatankhah-V、W. F. M. Daniel、Q. Li、A. P. Zhushma、A. V. Dobrynin、S. S. Sheiko，Nat.Commun.2016, 7, 12919.

[17] C. Yu, H. Guo, K. Cui, X. Li, Y. N. Ye, T. Kurokawa, J. P. Gong, Proc. Natl. Acad. Sci. USA 2020, 117, 18962.
[17] C. Yu、H. Guo、K. Cui、X. Li、Y. N. Ye、T. Kurokawa、J. P. Gong，Proc.Natl.USA 2020, 117, 18962.

[18] X. Le, H. Shang, H. Yan, J. Zhang, W. Lu, M. Liu, L. Wang, G. Lu, Q. Xue, T. Chen, Angew. Chem., Int. Ed. 2021, 60, 3640.
[18] X. Le，H. Shang，H. Yan，J. Zhang，W. Lu，M. Liu，L. Wang，G. Lu，Q. Xue，T. Chen，Angew.Chem.Ed.2021, 60, 3640.

[19] Y. Liu, J. K. Boyles, J. Genzer, M. D. Dickey, Soft Matter 2012, 8, 1764

[20] Y. Liu, B. Shaw, M. D. Dickey, J. Genzer, Sci. Adv. 2017, 3, e1602417.

[21] G. Zhang, W. Peng, J. Wu, Q. Zhao, T. Xie, Nat. Commun. 2018, 9, 4002.
[21] G. Zhang, W. Peng, J. Wu, Q. Zhao, T. Xie, Nat.Commun.2018, 9, 4002.
[22] R. P. Sijbesma, F. H. Beijer, L. Brunsveld, B. J. B. Folmer, J. H. K. Ky Hirschberg, R. F. M. Lange, J. K. L. Lowe, E. W. Meijer, Science 1997, .
[22] R. P. Sijbesma, F. H. Beijer, L. Brunsveld, B. J. B. Folmer, J. H. K. Ky Hirschberg, R. F. M. Lange, J. K. L. Lowe, E. W. Meijer, Science 1997, .

[23] J. Li, J. A. Viveros, M. H. Wrue, M. Anthamatten, Adv. Mater. 2007, 19, 2851.
[23] J. Li、J. A. Viveros、M. H. Wrue、M. Anthamatten，Adv. Mater.2007, 19, 2851.

[24] G. Zhang, Q. Zhao, W. Zou, Y. Luo, T. Xie, Adv. Funct. Mater. 2016, .
[24] G. Zhang, Q. Zhao, W. Zou, Y. Luo, T. Xie, Adv.Funct.Mater.2016, .

[25] Y. Meng, W. Xu, M. R. Newman, D. S. W. Benoit, M. Anthamatten, Adv. Funct. Mater. 2019, 29, 1903721.
[25] Y. Meng, W. Xu, M. R. Newman, D. S. W. Benoit, M. Anthamatten, Adv.Funct.Mater.2019, 29, 1903721.

[26] C. E. Hoyle, C. N. Bowman, Angew. Chem., Int. Ed. 2010, 49, 1540.
[26] C. E. Hoyle, C. N. Bowman, Angew.Chem.Ed.2010, 49, 1540.

[27] W. P. Appel, G. Portale, E. Wisse, P. Y. Dankers, E. W. Meijer, Macromolecules 2011, 44, 6776.

[28] C. Liu, Z. Fan, Y. Tan, F. Fan, H. Xu, Adv. Mater. 2020, 32, 1907569.
[28] C. Liu, Z. Fan, Y. Tan, F. Fan, H. Xu, Adv. Mater.2020, 32, 1907569.

[29] S. Choi, Y. Eom, S.-M. Kim, D.-W. Jeong, J. Han, J. M. Koo, S. Y. Hwang, J. Park, D. X. Oh, Adv. Mater. 2020, 32, 1907064.
[29] S. Choi、Y. Eom、S.-M.Kim、D.-W.Jeong, J. Han, J. M. Koo, S. Y. Hwang, J. Park, D. X. Oh, Adv. Mater.2020, 32, 1907064.

[30] L. Romano, A. Portone, M.-B. Coltelli, F. Patti, R. Saija, M. A. Iatì, G. Gallone, A. Lazzeri, S. Danti, O. M. Maragò, A. Camposeo, D. Pisignano, L. Persano, Nat. Commun. 2020, 11, 5991.
[30] L. Romano, A. Portone, M.-B.Coltelli, F. Patti, R. Saija, M. A. Iatì, G. Gallone, A. Lazzeri, S. Danti, O. M. Maragò, A. Camposeo, D. Pisignano, L. Persano, Nat.Commun.2020, 11, 5991.

[31] Y. Mao, K. Yu, M. S. Isakov, J. Wu, M. L. Dunn, Q. H. Jerry, Sci. Rep. .
[31] Y. Mao, K. Yu, M. S. Isakov, J. Wu, M. L. Dunn, Q. H. Jerry, Sci. Rep. .

[32] Q. Ge, A. H. Sakhaei, H. Lee, C. K. Dunn, N. X. Fang, M. L. Dunn, Sci. Rep. 2016, 6, 31110.

[33] B. Peng, Y. Yang, K. A. Cavicchi, Multifunct. Mater. 2020, 3, 042002.
[33] B. Peng, Y. Yang, K. A. Cavicchi, Multifunct.Mater.2020, 3, 042002.

[34] Y. Foelen, D. A. Heijden, M. Pozo, J. Lub, C. W. Bastiaansen, A. P. Schenning, ACS Appl. Mater. Interfaces 2020, 12, 16896.
[34] Y. Foelen, D. A. Heijden, M. Pozo, J. Lub, C. W. Bastiaansen, A. P. Schenning, ACS Appl.Interfaces 2020, 12, 16896.