In Situ Time-Resolved Attenuated Total Reflectance Infrared Spectroscopy for Probing Metal-Organic Framework Thin Film Growth

Abstract

In situ chemical measurements of solution/surface reactions during metal-organic framework (MOF) thin film growth can provide valuable information about the mechanistic and kinetic aspects of key reaction steps, and allow control over crystal quality and material properties. Here, we report a new approach to study the growth of MOF thin films in a flow cell using attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR). Real-time spectra recorded during continuous flow synthesis were used to investigate the mechanism and kinetics that govern the formation of (Zn, Cu) hydroxy double salts (HDSs) from ZnO thin films and the subsequent conversion of HDS to HKUST-1. We found that both reactions follow pseudo-first order kinetics. Real-time measurements also revealed that the limited mass transport of reactants may lead to partial conversion of ZnO to HDS and therefore leaves an interfacial ZnO layer beneath the HDS film providing strong adhesion of the HKUST-1 coating to the substrate. This in situ flow-cell ATR-FTIR method is generalizable for studying the dynamic processes of MOF thin film growth, and could be used for other solid/liquid reaction systems involving thin films.

Graphical abstract

Introduction

With their intrinsic porosity and a diversity of crystalline structures, metal-organic frameworks (MOFs) have attracted tremendous interest in the past few decades.^1,2 Over 20 000 MOF structures have been reported,¹ and these materials have been investigated for a wide array of applications, such as gas adsorption, separations, and catalysis.^3–7 However, due to the difficulty in probing crystal growth in solutions,⁸ obtaining direct evidence to elucidate MOF growth mechanisms continues to be a challenge.⁹ As a result, dynamic processes that take place during MOF formation are still not well understood, and MOF synthesis often relies on trial and error.⁸ Understanding the mechanism of MOF nucleation and growth will enhance control over the material properties, morphology and defects,¹⁰ and allow the discovery of improved synthesis methods.⁹

Several in situ measurements have been previously reported to study how MOFs form, including atomic force microscopy (AFM),^10,11 surface plasmon resonance (SPR) spectroscopy,¹² quartz crystal microbalance (QCM),¹³ X-ray diffraction (XRD),^14,15 small-angle and wide-angle X-ray scattering (SAXS/WAXS),^16–18 and liquid cell transmission electron microscopy (LCTEM).¹⁹ Attfield et al. probed the surfaces of MOF crystals using in situ AFM during the growth of HKUST-1 and MOF-5.^10,11 With sub-nanometer resolution, this technique can track the growth of crystal surfaces and identify defects present during crystallization. In situ SPR and QCM are powerful tools to analyze the kinetics for the formation of surface-mounted MOFs in real time.^12,13 In situ XRD and SAXS/WAXS have been used to follow the transition of crystal structures during MOF synthesis and identify intermediate phases.^14–18 Recently, Patterson et al. captured the dynamic processes during ZIF-8 growth using LCTEM, providing direct visual evidence for ZIF-8 formation from small constituents.¹⁹

While previous in situ methods have been mainly focused on the changes in crystal structure, crystal morphology, crystal size, and mass uptake on surfaces, these measurements do not directly reveal chemical information such as bond formation and ion/ligand exchange during the process of MOF growth. Here, we report for the first time the use of in situ time-resolved attenuated total reflectance (ATR) Fourier transform infrared (FTIR) spectroscopy in a continuous-flow reactor to study the mechanism and kinetics of MOF growth. The ATR technique relies on the total reflection of the IR beam in an internal reflection element (IRE), where surface enhancement is achieved by the evanescent wave formed perpendicular to the reflecting surface in the IRE.^20,21 Compared to transmission FTIR spectroscopy, ATR-FTIR is advantageous for probing surfaces (rather than measuring the bulk) and particularly powerful for in situ analyses of chemisorption and reactions at solid/liquid interfaces when the liquid is highly absorbing.²¹ Although liquid-cell ATR measurements are commonly employed for the investigation of a wide variety of molecular adsorption and surface reactions,^22–28 ATR-FTIR has not yet been explored for real-time monitoring of MOF growth.

Key challenges of using ATR-FTIR to study MOF growth are: i) the limited penetration depth of the evanescent wave to probe through the reagent solution; and ii) the lack of controlled MOF nucleation and growth on the IRE. For example, during common solvothermal syntheses, MOF crystals formed in the solution are often observed to fall onto the bottom of the reaction vessel. Consequently, it is difficult to distinguish between homogeneous nucleation (in the solution) and heterogeneous nucleation (on the IRE surface). Once the IRE is covered with a sufficiently thick layer (greater than 1 μm thick) of MOF crystals, the ATR configuration will no longer be able to probe the extent of reaction. After the IRE surface is overcoated with MOFs, it needs to be cleaned or replaced to carry out additional measurements.

To deal with these challenges, we designed an ATR configuration to use 50 mm diameter Si wafers as the IRE as shown in Figure 1a (a detailed description of the ATR-FTIR setup is provided in Figure S1). This approach is advantageous because Si wafers are replaceable and less costly than common IREs (e.g. trapezoidal prisms and other geometries). As a test-case for our measurement, we focus on the recently reported processes that show controlled MOF nucleation and growth on or from surface metal oxides.^29–36 We use atomic layer deposition (ALD) to prepare a metal oxide thin film on the Si wafer IRE as the starting material or nucleation layer for MOF growth (Figure 1b). In this work, we will use our custom-built flow-cell ATR-FTIR to study the mechanism and kinetics of a recently reported conversion of hydroxy double salt (HDS) to MOFs.³³ In this two-step scheme, ZnO reacts with Cu(NO₃)₂ in solution to form (Zn, Cu) hydroxy nitrate, and this HDS intermediate is further converted to HKUST-1 MOF via anion exchange at room temperature within 1 min (Figure 1c). Previous work also confirms that this approach is generally applicable to other systems, including for example, Cu-BDC, IRMOF-3, ZIF-8 and other MOF materials.³³ Ex situ characterization has been used in the previous report to investigate the reaction mechanism.³³ Here, we show that in situ flow-cell ATR-FTIR can capture the dynamics of HDS formation and subsequent anion exchange in the HDS layer that ultimately leads to MOF formation.

Figure 1.

(a) Schematic of the in situ ATR-FTIR spectroscopy setup equipped with a liquid flow cell. An undoped silicon wafer is used as the internal reflection element. (b) Schematic of the solid-liquid interface for the surface reactions to form MOF thin films. A layer of ZnO deposited on Si using ALD is used as a templating material for rapid room-temperature synthesis of the HKUST-1 thin film.

Results and Discussion

We deposited 300 cycles of ALD ZnO (≈ 60 nm thickness) on the Si wafer as the initiation layer for HDS formation. In a typical room-temperature flow synthesis experiment, 300 mmol/L of Cu(NO₃)₂ solution in ethanol-water (ethanol volume fraction of 0.5) was first pumped into the liquid cell for 120 s, followed by a rinse step using the mixed solvent for 60 s. Subsequently, 10 mmol/L of 1,3,5-benzenetricarboxylic acid (H₃BTC) solution in ethanol-water was pumped into the cell for 300 s. After the flow of H₃BTC solution, the cell was further rinsed with the ethanol-water solvent mixture for 120 s. The flow rate was controlled at 100 mL/min during the entire process. We modified the previously published procedure³³ to exclude dimethylformamide (DMF) from the solvent mixture, because the strong carbonyl band in DMF at ≈ 1680 cm⁻¹ was found to overlap with the asymmetric stretching mode for the carboxylate groups (v_as(OCO⁻)) in HKUST-1. Infrared absorbance spectra were recorded every 0.49 s and each spectrum was referenced against steady-state flow of the solvent mixture in the absence of reactants. Sets of spectra were analyzed by integrating the peaks for $v({\text{NO}}_3^{-})$ at ≈ 1420 cm⁻¹, v(C=C) at ≈ 1588 cm⁻¹ and v_as (OCO⁻) at ≈ 1647 cm⁻¹.³³ We used peak areas to follow the extent of reaction during HDS formation from ALD ZnO as well as HDS conversion to HKUST-1.

Figure 2a shows time-resolved ATR-FTIR spectra for the surface reactions to form HKUST-1 thin films (corresponding 3D plot is shown in Figure S2). Two broad peaks at ≈ 1360 cm⁻¹ and ≈ 1420 cm⁻¹ appear after the ZnO coated IRE is exposed to Cu(NO₃)₂ solution in the flow cell. These two peaks are associated with vibrations of the ${\text{NO}}_3^{-}$ groups mainly present in the (Zn, Cu) hydroxy nitrate double salt.³³ The integrated peak area for $v({\text{NO}}_3^{-})$ at ≈ 1420 cm⁻¹ (Figure 2b) increases dramatically and reaches saturation after 1 min, indicating the fast formation of (Zn,Cu) HDS from the ZnO thin film. Following the exposure of the ZnO surface to Cu(NO₃)₂ solution, ethanol-water solvent mixture was used to rinse the cell for 1 min. The absorbance of the $v({\text{NO}}_3^{-})$ peaks remains constant during the rinse step. When the H₃BTC solution is dosed into the flow cell at t ≈ 3.5 min, four peaks appear (Figure 2a) which represent the symmetric and asymmetric stretching modes (1378 cm⁻¹ and 1647 cm⁻¹, respectively) of the carboxylate groups in HKUST-1,³³ and the C=C vibration modes (1450 cm⁻¹ and 1588 cm⁻¹) associated with the aromatic ring in the BTC³⁻. The decreased peak area of $v({\text{NO}}_3^{-})$ and the increased peak area of v_as(OCO⁻) and v(C=C) were observed simultaneously (Figures 2b–d), consistent with the anion exchange process reported for HDS conversion to MOFs.³³ The v_as(OCO⁻) and v(C=C) peaks reach their maximum area within 1 min, showing the rapid conversion of HDS to HKUST-1. SEM images and XRD patterns in the supporting information (Figure S3 and S4) confirm that the film formed on the surface is HKUST-1 MOF with crystal sizes of 571 nm ± 124 nm.

Figure 2.

(a) Time-resolved ATR-FTIR spectra for surface reactions to form HKUST-1 thin films. (b–d) Integrated area for IR peaks representing $v({\text{NO}}_3^{-})$ at ≈ 1420 cm⁻¹, v(C=C) at ≈ 1588 cm⁻¹, and v_as(OCO⁻) at ≈ 1647 cm⁻¹.

Since the flow cell ATR-FTIR measurement can follow the extent of surface reactions, we investigated the kinetics for the formation of HDS from ZnO and the conversion of HDS to HKUST-1. Figure 3a shows the change of integrated IR peak area for $v({\text{NO}}_3^{-})$ at ≈ 1420 cm⁻¹ as a function of reaction time when the initial ZnO layer (≈ 60 nm) is reacted with different concentrations of Cu(NO₃)₂ solution. For all concentrations tested for this reaction, $v({\text{NO}}_3^{-})$ peak area rises rapidly in the first 30 s of reaction and then slowly saturates as reaction reaches steady-state conditions. In order to correlate the IR absorbance (A) with the concentration of nitrate groups, we derived Equation 1 (full derivation provided in the SI) based on the differential form of the Beer-Lambert law and the exponential decay of the electric field strength for the evanescent wave.

Figure 3.

(a) Integrated IR peak area for $v({\text{NO}}_3^{-})$ at ≈ 1420 cm⁻¹ as a function of reaction time during the flow of Cu(NO₃)₂ solutions with different concentrations (75 mmol/L to 600 mmol/L) on a fixed thickness of ALD ZnO (≈ 60 nm). (b) Initial slope (0 ≤ t ≤ 0.1 min) derived from the curves in (a). The dashed curve represents the fit to Equation 5. (c) Integrated $v({\text{NO}}_3^{-})$ peak area during the flow of 300 mmol/L Cu(NO₃)₂ on different thicknesses of ALD ZnO layers (≈ 20 nm to ≈ 100 nm). (d) Integrated $v({\text{NO}}_3^{-})$peak area after HDS growth saturation (t = 2.0 min). Error bars in (b) and (d) represent standard deviation for each sample.

$$A=\frac{N\alpha }{ln10}\left[\frac{C_{N,\mathit{Sol}}}{2\gamma }+(C_{N,\mathit{HDS}}-C_{N,\mathit{Sol}})L_{\mathit{HDS}}\right]$$ (1)

where N is the number of total reflections, C_{N, HDS} and C_{N, Sol} (in units of mmol/L) are the concentrations of nitrate groups in the HDS film and in the solution, respectively. C_{N, HDS} and C_{N, Sol} are both considered constant during the initial period of reaction. L_HDS is the film thickness of (Zn, Cu) HDS and α, γ are defined in Equation 2 and Equation 3, respectively.

$$\alpha =\epsilon I_0{e}^{-2\gamma d_{\mathit{Zno}}}$$ (2)

where ε is the molar attenuation coefficient for the nitrate groups (in units of m²/mol), I₀ is the intensity of the incident evanescent wave at the IRE surface, d_ZnO is the distance from the IRE surface to the top surface of the ZnO film. γ is defined as the reciprocal of the penetration depth (d_p), and depends on the refractive indices of the IRE (n₁) and the thin film on top of the IRE (n₂), the angle of incidence (θ), and the IR wavelength (λ).^37,38

$$\gamma =\frac{1}{d_p}=\frac{2\pi n_1\sqrt{{\mathit{sin}}^2\theta -{\left(\frac{n_2}{n_1}\right)}^2}}{\lambda }$$ (3)

The chemical reactions possibly involved in the process to form HDS from ZnO are written as Reaction I and II.

$$\mathit{ZnO}+2H^{+}\to Z{n}^{2+}+H_{2}O$$ (I)

$$Z{n}^{2+}+xC{u}^{2+}+(y+z)H_{2}O+(2+2x-y)N{O}_3^{-}\rightleftharpoons \mathit{ZnC}{u}_x{(OH)}_y{(N{O}_3)}_{2+2x-y}\cdot z{H}_{2}O+y{H}^{+}$$ (II)

Since it is non-trivial to identify all the elementary reactions and to determine a rate limiting step, we assume that the entire conversion process from ZnO to HDS follows pseudo-first order kinetics (Equation 4).

$$-r_1=k_{\mathit{app}}\cdot C_{cu}$$ (4)

where r₁ is the reaction rate for the overall reaction process, k_app is the apparent first order rate constant, C_Cu is the concentration of Cu²⁺. Equation 1 then becomes

$$A=\frac{N\alpha }{ln10}\cdot \frac{C_{N,\mathit{Sol}}}{2\gamma }+\frac{(C_{N,\mathit{HDS}}-C_{N,\mathit{Sol}})C_{Cu}}{C_{N,\mathit{HDS}}}\cdot \phi t$$ (5)

where t is the reaction time (in units of min). The grouped kinetic parameter φ is given as

$$\phi =\frac{N\alpha }{ln10}\cdot \frac{k_{\mathit{app}}V}{S}$$ (6)

where V (m³) is the volume of Cu(NO₃)₂ solution involved in the reaction to form the HDS, and S (m²) is the geometric area of the HDS film, i.e. the planar surface area on the wafer exposed to reactants.

As indicated in Equation 5, absorbance is linearly related with the reaction time. This allows us to interpret the initial slope (0 ≤ t ≤ 0.1 min) of the curves in Figure 3a as an indication of reaction rate. Figure 3b shows that the initial slope increases with the Cu(NO₃)₂ concentration. Upon fitting the data with Equation 5, we find that the experimental results agree well with the model, confirming that the ZnO-to-HDS conversion can be described by pseudo-first order kinetics. Fitting of the initial slope using Equation 7 gives φ = (0.041 ± 0.004) m³∙mol⁻¹∙min⁻¹.

$$\mathit{Inital}\mathit{Slope}=\frac{(C_{N,\mathit{HDS}}-C_{N,\mathit{Sol}})C_{Cu}}{C_{N,\mathit{HDS}}}\cdot \phi$$ (7)

We also studied the effect of the starting thickness of the ZnO films on the kinetics of HDS formation. Figure 3c shows the integrated $v({\text{NO}}_3^{-})$ peak area during the flow of 300 mmol/L Cu(NO₃)₂ solutions on different thicknesses of ALD ZnO layers (100 cycles to 500 cycles, ≈ 20 nm to ≈100 nm). As expected, the initial slope of the curves in Figure 3c are comparable for all oxide thicknesses tested. To compare the amount of HDS formed at the end of Cu(NO₃)₂ flow, we plotted the final integrated $v({\text{NO}}_3^{-})$ peak area at t = 2 min as a function of the thicknesses of the ZnO films deposited on the Si IRE (Figure 3d). While the $v({\text{NO}}_3^{-})$ peak area shows a dependence on the initial ZnO thickness, for ZnO films ≥ 60 nm in thickness, $v({\text{NO}}_3^{-})$ peak area exhibits saturation. We calculated the d_p of the ATR evanescent wave based on Equation 3 and compared this value with the HDS film thicknesses measured from cross sectional SEM images. In the wavenumber range (1250 cm⁻¹ to 1750 cm⁻¹) used to track the reaction that forms (Zn, Cu) HDS from ZnO, d_p ranges from 0.62 μm to 0.88 μm. These d_p values are larger than the thickest HDS layer (0.56 μm) converted from ≈ 100 nm thick ALD ZnO (Figure S5), confirming that surface changes during the ZnO-to-HDS conversion are within the information depth of the ATR-FTIR measurement. Transmission IR measurements also indicate that the film thicknesses of the HDS converted from thick ZnO layers (≥ 60 nm) are indeed very similar (Figure S6). These results reveal that full conversion is not achieved for thick ZnO starting layers during flow synthesis.

The partial conversion of thick ZnO layers to (Zn, Cu) HDS possibly stems from a diffusion-limited process. As shown in Figure 3c, for ZnO initial layers with thicknesses ≥ 60 nm, the rise in the $v({\text{NO}}_3^{-})$ peak area stops after 0.5 min. The dense HDS formed on the surface may impede reactant diffusion into the film, thus hindering mass transport (Figure 4). Consequently, the growth of HDS stops at a certain depth where the reactant is either depleted or is low in concentration, leaving an interfacial layer of ZnO in between the HDS film and the Si substrate. The XRD results in Figure S4 also confirm the existence of ZnO in the HDS films synthesized from thick ZnO initiation layers.

Figure 4.

Schematic illustration of the reaction and diffusion processes for thick and thin ZnO starting layers during the formation of HDS and its subsequent conversion to HKUST-1.

We observed that the ZnO interfacial layer due to partial conversion to HDS plays a critical role in attaching the MOF thin films to the substrate during the conversion of as-synthesized (Zn, Cu) HDS to HKUST-1 films (Figure 4). To demonstrate this, we studied the formation of HKUST-1 films using 10 mmol/L of H₃BTC solution and HDS prepared from different thicknesses of ZnO initial films. Time-resolved IR spectra confirm the formation of HKUST-1 from (Zn, Cu) HDS regardless of the thickness of the ZnO initial layers (Figure S7). In the cases where a ZnO interfacial layer remained beneath the MOF films (for ZnO thicknesses ≥ 60 nm), dense and uniform HKUST-1 films were obtained and found to be attached to the substrate as shown in the SEM images in Figure S8. In comparison, when a thin ZnO initial layer was used, the complete consumption of the ZnO film and the subsequent conversion of HDS into MOF resulted in the delamination and loss of the MOF layer.

We further investigated the kinetics for the conversion of (Zn, Cu) HDS to HKUST-1 by monitoring the v_as(OCO⁻) peak at ≈ 1647 cm⁻¹. HDS thin films with and without ZnO interfacial layers were synthesized from ≈ 60 nm and ≈ 20 nm thick ALD ZnO films, respectively. The concentration of H₃BTC solution used in the experiments ranged from 1 mmol/L to 20 mmol/L. Figure 5a shows the formation of HKUST-1 from HDS with the presence of a ZnO interfacial layer. The area of the v_as(OCO⁻) peak increases dramatically and then reaches a plateau which indicates maximum conversion of HDS. In comparison, the peak area of v_as(OCO⁻) shown in Figure 5b drops after initial growth, indicating film delamination and loss during reactant flow in the cases without the ZnO interfacial layers. Indeed, we observed sparse islands of HKUST-1 crystals on these substrates by SEM (Figure S8).

Figure 5.

(a–b) Integrated IR peak area for v_as(OCO⁻) at ≈ 1647 cm⁻¹ as a function of reaction time during the flow of H₃BTC solutions on (Zn, Cu) HDSs converted from (a) ≈ 60 nm and (b) ≈ 20 nm of ALD ZnO thin films. (c) Initial slope (0 ≤ t ≤ 0.1 min) de-rived from the curves in (a–b) as a function of H₃BTC concentration. The dashed line represents linear fitting. (d) Integrated IR peak area for v_as(OCO⁻) at ≈ 1647 cm⁻¹ at t = 6 min, representing the final amount of HKUST-1 synthesized from HDS thin films with and without a ZnO interfacial layer. Error bars in (c) and (d) represent standard deviation for each sample.

The reaction associated with the HDS-to-MOF conversion is given in Reaction III.

$$\mathit{ZnC}{u}_x{(OH)}_y{(N{O}_3)}_{2+2x-y}\cdot z{H}_{2}O+\frac{2x}{3}{H}_3\mathit{BTC}\to \frac{x}{3}C{u}_{3}BT{C}_2\cdot n{H}_{2}O+Z{n}^{2+}+(2x-y)H^{+}+(2+2x-y)N{O}_3^{-}+\left(2x-y+z-\frac{nx}{3}\right)H_{2}O$$ (III)

With a derivation similar to Equation 1 and maintaining the assumption of pseudo-first order kinetics, we obtain Equation 8 (full derivation provided in SI) to correlate absorbance A′ for v_as(OCO⁻) with reaction time.

$$A^{\prime }={\phi }^{\prime }{C}_{\mathit{BTC}}t$$ (8)

where the grouped kinetic parameter φ′ for the HDS-to-MOF conversion is defined in Equation 9. C_BTC is the concentration of H₃BTC in the solution.

$${\phi }^{\prime }=\frac{N{\alpha }^{\prime }}{ln10}\frac{k_{\mathit{app}}^{\prime }{V}^{\prime }}{S^{\prime }}$$ (9)

where α′, $k_{\mathit{app}}^{\prime }$, V′ and S′ are similar to their counterparts in Equations 2 through 4, and defined in the SI.

Equation 8 establishes a linear relationship between absorbance and reaction time, so we can use the initial slope of the integrated v_as(OCO⁻) curve to obtain φ′. The initial slope (0 ≤ t ≤ 0.1 min) obtained from the curves in Figure 5a–b exhibits a linear dependence on H₃BTC concentration (Figure 5c). This is consistent with our model, and confirms that the conversion of HDS to HKUST-1 is also described by pseudo-first order kinetics. φ′ obtained from linear fitting gives a value of (0.338 ± 0.004) m³∙mol⁻¹∙min⁻¹ for the initial reaction on HDS layers synthesized from ≈ 60 nm thick ZnO, and (0.385 ± 0.017) m³∙mol⁻¹∙min⁻¹ for initial MOF formation on HDS converted from ≈ 20 nm thick ZnO films. The difference between these two φ′ values could be a result of different evanescent wave intensities at the top surfaces of thick and thin HDS films (Equation S4).

Figure 5d shows the integrated area for the v_as(OCO⁻) peak after the termination of H₃BTC flow, which represents the total amount of HKUST-1 remaining on the IRE surface. For HDS films obtained from ≈ 60 nm thick ZnO, the as-synthesized MOF layers were all found to be attached to the substrate regardless of the H₃BTC concentration. Moreover, the integrated v_as(OCO⁻) area increases as a function of H₃BTC concentration. Compared with other concentrations tested, 1 mmol/L of H₃BTC results in notably lower conversion of the HDS layer synthesized from ≈ 60 nm thick ZnO. This may be because the mass transport rate for H₃BTC to diffuse into the HDS film is low due to the relatively small concentration gradient, while the reaction rate to form HKUST-1 with 1 mmol/L of H₃BTC is high. The diffusion of H₃BTC in the HKUST-1 growth front could be a secondary factor affecting HDS conversion, depending on the relative diffusivities of H₃BTC in HDS and HKUST-1, respectively. Consequently, the diffusion-controlled process leads to limited conversion of the HDS film.

For HDS films grown from a thin ZnO layer (≈ 20 nm), the final integrated v_as(OCO⁻) area for the reaction run with 1 mmol/L H₃BTC is significantly higher than reactions run with ≥ 2 mmol/L H₃BTC (Figure 5d). The lower effective conversion at increasing H₃BTC concentrations can be explained by rapid delamination of the MOF layer. Therefore, when the H₃BTC concentration is high and the starting ZnO layer is relatively thin, the reaction front quickly reaches the HDS/Si interface and results in the detachment of the MOF film. These results also indicate that film loss could be controlled by using low concentrations of H₃BTC solution. Alternatively, limiting reaction time before delamination occurs when thin HDS films are to be employed (based on Figure 5b) could also be effective for achieving densely-covered thin (e.g., < 100 nm) HKUST-1 films using HDS conversion.

Conclusion

We have demonstrated a custom-designed flow-cell ATR-FTIR system capable of probing surface reactions in situ during MOF thin film synthesis. This system was used to investigate the conversion of ZnO to (Zn, Cu) HDS and the subsequent conversion of HDS to HKUST-1 MOF at room temperature. Time-revolved spectra show that both ZnO-to-HDS and HDS-to-MOF reactions reach maximum conversion in under 1 min. Combined experiment and modeling analyses reveal pseudo-first order kinetics for each conversion step. The thickness of the HDS and the resulting MOF film product is found to be tunable by adjusting the thickness of ALD ZnO initial layers. We found that partial conversion of the initial ZnO film due to diffusion-limited reaction leads to an interfa-cial ZnO layer that can enhance the adhesion of the MOF film to the substrate. Our results provide important insights into the reaction conditions needed to produce surface-bound ultra-thin MOF films. The flow reactor system with in situ ATR-FTIR described here is expected to aid the development and analysis of other thin film deposition processes as well.