**2. Results and Discussion**

In this work, four azo ligands with different lengths of alkyl linkers and substituent groups were covalently grafted onto the lacunary Keggin of [SiW11O39] <sup>8</sup><sup>−</sup> (SiW11) (denoted as azo-Keggins). The as-prepared azo-Keggins showed the general molecular formula of [C16H36N]4[SiW11O40(Si(CH2)3–NH–CO(CH2)nO–C6H4N=NC6H4–R)2] (R = H, *n* = 0 (**1**); R = NO2, *n* =0(**2**); R = H, *n* =5(**3**); R = H, *n* = 10 (**4**)). Different from the previously reported azo-Andersons, in which two azo ligands were attached to the opposite sides of MnMo6O18, the resultant azo-Keggins contained two azo ligands on the same sides of SiW11 (Figure 1). Considering strong intermolecular interactions was the prerequisite for the formation of large aggregates in the solution or gas phase; these azo-Keggins were expected to have weak intermolecular interactions due to hindrance from Keggin clusters with large molecular size and strong electrostatic repulsion.

The azo-POMs can be detected by electrospray ionization mass spectrometry (ESI-MS), as the aggregates commonly yielded multiple species with similar *m*/*z* resulting in overlapping envelopes in the mass spectra [18]. The ESI-MS spectra of compounds **1**–**4** were acquired by directly using their acetonitrile solutions. In each case, a series of notable peaks was observed and can be well assigned to the corresponding fragment ions, demonstrating that these hybrid structures were successfully prepared and remained intact both in the solution and gas phase. As shown in Figure 2, compounds **1**–**4** showed MS signals at *m*/*z* 1898.1, 1943.0, 1968.1, and 2038.2, which can be assigned to the fragment ions of [**X1**–**4**+2TBA]2- (**X1**–**<sup>4</sup>** = the anionic part of compounds **1**–**4**; TBA = tetrabutylammonium). All four peaks provided the unambiguous isotopic distribution envelopes without any

similar overlapping fashion, suggesting the monomeric state of azo-Keggins. Full spectra and a list of identified peaks are provided in the Figures S1–S4.

**Figure 2.** ESI-MS spectra of the compound **1**–**4** (red peaks in **a**–**d**) showing the *m/z* towards fragment ions of [**X1**–**<sup>4</sup>** + 2TBA]2<sup>−</sup> (**X1–4** = anionic part of **1–4**). The corresponding simulated isotopic patterns of the [**X1–4** +2TBA]2<sup>−</sup> were indicated as the grey peaks in **a**–**d**.

After confirming the monomeric state of azo-Keggins with intact molecular structure, IMS/MS measurements were applied to analyze their bistability conformation and reversible dynamics, considering that MS was unable to provide the information on shape. A commercially available travelling-wave ion-mobility spectrometer (TWIMS) was coupled with mass measurements to separate the ions according to their mobility and to derive the collision cross-section (CCS) data of the ions. Figure 3 showed the 2D IMS/MS spectrum of compound **1**, with the peak intensity displayed by a color-coded logarithmic scale. The *x*-axis represented the *m*/*z* range from MS, and the *y*-axis represented the drift time from IMS. As we might expect, the spectrum of compound **1** showed a clearer situation than that of azo-Andersons, where larger oligomeric structures or the higher charge states were minor. Same results were observed in the 2D IMS/MS spectra of **2**–**4** both in their *trans*and *cis*-conformations (Figures S5–S12). The intense (yellow) line of the peaks could be easily assigned to the individual cluster ions as observed in ESI-MS. All these peak envelopes could be assigned. The intensive peak at *m/z* = 1989.1 and drift time = 10.8 ms was assigned to the monomeric Keggin hybrids: [**X1** + 2TBA]2<sup>−</sup> (**X1** = anionic part of **1**). The corresponding higher charge peaks were assigned to minor aggregates: [**2X1** + 5TBA]3<sup>−</sup>, [**3X1**+8TBA]4<sup>−</sup>, and [**4X1**+11TBA]5<sup>−</sup> (Figure 3).

**Figure 3.** The 2D IMS/MS spectrum of compound **1**. The diagonal lines of similarly charged species were encircled by ellipsoids and the charges of these species were given in white. The peak in green rectangle was assigned to [**X1** + 2TBA]2<sup>−</sup>.

With the further inspection of the IMS/MS spectrum of compound **1**, the conformational variation of these azo functionalized POM macromolecules could be observed. It was well known that the conformation of the azo bond can change from *trans* to *cis* upon UV irradiation at 365 nm and reversibly recovered back to *trans* conformation when the sample was exposed to visible light. The two conformations were relatively stable under the constant light condition, which enabled the detection of the bistability of azo-Keggins and the reversible transformation process. Additional UV-Vis and 1H NMR experiments were conducted before and after UV irradiation at 365 nm. Moreover, the kinetic curve of the UV absorbance with time increasing was provided to follow the isomerization process (Figure S20). The results showed that almost all *trans*-isomers were changed to *cis*-isomers after 5 h UV light irradiation. The *cis*-isomer samples were immediately measured by IMS/MS under the exclusion of light. Figure 4 showed the drift time spectra for the main MS peaks of anionic part of compounds **1**–**4** with two TBA cations [**X1–4** + 2TBA]2−. Before and after UV irradiation, the drift time peaks (as indicated by green rectangle in Figure 3) manifested a clear shift, suggesting that the shape of compounds was changed. In IMS, the drift velocity of the ions was proportional to the electric field, and proportionality constant *K* was related to the CCS of ions [23]. On the basis of the CCS value that can be measured directly from the drift time, information about the chemical structure and 3D conformation of the ions was further provided (Tables S1–S4 Supplementary Materials).

Furthermore, the stability of specific isomers can be reflected in the intensity of drift time peaks. As shown in Figure 4, **1-UV** showed a lower peak intensity than its isomer **1-Vis**, suggesting the lower stability of *cis* conformation than that of *trans* conformation. The decreased stability was attributed to the linkage of the Keggin POM cluster, which was considered a strong electron-withdrawing group [35]. Comparing the peak intensity of **2-UV** and **2-Vis**, the terminal substituent of electron-withdrawing NO2 groups further decreased the stability of *cis* conformation [36]. Similar peak intensities were observed between two conformational isomers toward **3** and **4**, suggesting that the effect of substituent on the stability of isomers was diminishing with the increased length of alkyl chains between POMs and azo groups. Therefore, the conformation bistability of compounds **3** and **4** could be achieved.

**Figure 4.** Comparison of the drift time graphs of the main MS peaks of compounds **1**–**4** with visible (solid lines) and UV (dash lines) irradiation. The spectra were consistent for all three repeated cycles.

As shown in Figure 5, the CCS value between the *trans* and *cis* conformations of compounds **1**–**4** showed the difference ranging from 6.91 Å2 for the linker with the longest alkyl chain to 2.44 Å<sup>2</sup> for the smallest compound. Compared with the dimeric Anderson azo-POMs (CCS difference from 26 Å<sup>2</sup> to 13 Å2), monomeric azo-Keggins exhibited a higher resolution difference, which could more precisely reflect the subtle conformational change of macromolecules derived from the azo group itself. The CCS differences for compounds **1**–**2** were 2.44 Å<sup>2</sup> and 5.07 Å2, respectively, which were consistent with both of the roughly calculated conformational difference of the azo or nitro-azo groups by ChemBio3D Software and 2.1 Å2 CCS difference of the conformational variation of the protonated azo reported in the literature [35]. As for compounds **3**–**4**, the greater CCS difference value was attributed to the more significant shape variation caused by flexible alkane chains. Therefore, for macromolecular azo-Keggins, the conformational change originated mainly from the transformation of organic azo ligands, while the influence from bulky POM clusters was minor. The IMS/MS measurement toward compounds **1**–**4** with reversible UV/Vis irradiation was repeated three times, and the resulting spectra were consistent through all repetitions of the isomerization, suggesting reproducibility.

It is worthwhile to note that our previous work reported the investigation of the IMSMS study of azo-Anderson assemblies [18]. In contrast, the azo-Keggin assemblies reported herein showed very different results: (1) The resultant azo-Keggins could stabilize the monomeric state, while azo-Anderson favored to form aggregates (dimer, trimer, and tetramer, etc.). Moreover, the conformational variation of azo-Keggins was more precisely reflected in IMS/MS results with a smaller CCS difference (2.44 Å2 to 6.91 Å2) than that of azo-Anderson aggregates with a large CCS difference (13 Å<sup>2</sup> to 26 Å2), which was due to the influence of the intermolecular arrangement. (2) The azo-Andersons with substituents of electron donor groups (such as alkoxyl chains) showed a higher stability of *cis-*conformation compared with that of a *trans-*conformation. As complementary, azo-Keggins with a substituent of electron-withdrawing groups exhibited a higher stability of *trans-*conformation than that of *cis-*conformation. When long alkyl chains were present in azo-Keggin assemblies (-C5 and -C10 alkyl chains were linked between azo and POMs), the bistability of both *trans-* and *cis-* isomers can be achieved.

**Figure 5.** The changes in collision cross-section of compounds **1**–**4** when switching between *trans* and *cis* conformations of the azo bond. The formula of the azo-Keggin monomers was [**X1–4** + 2TBA]2<sup>−</sup>.

#### **3. Materials and Methods**
