Analyte Detection with Cu-BTC Metal–Organic Framework Thin Films

Article pubs.acs.org/ac

Analyte Detection with Cu-BTC Metal−Organic Framework Thin Films by Means of Mass-Sensitive and Work-Function-Based Readout Polina Davydovskaya,*,† Annekatrin Ranft,‡,§,∥ Bettina V. Lotsch,‡,§,∥ and Roland Pohle† †

Siemens AG, Corporate Technology, Otto-Hahn-Ring 6, 81739 Munich, Germany Max Planck Institute for Solid State Research, Heisenbergstraße 1, 70569 Stuttgart, Germany § Department of Chemistry, University of Munich (LMU), Butenandtstraße 5-13, 81377 Munich, Germany ∥ Nanosystems Initiative Munich (NIM) and Center for Nanoscience, Schellingstraße 4, 80799 Munich, Germany ‡

S Supporting Information *

ABSTRACT: Metal−organic frameworks (MOFs) constitute a new generation of porous crystalline materials, which have recently come into focus as analyte-specific active elements in thin-film sensor devices. Cu-BTCalso known as HKUST-1 is one of the most theoretically and experimentally investigated members of the MOF family. Its capability to selectively adsorb different gas molecules renders this material a promising candidate for applications in chemical gas and vapor sensing. Here, we explore details of the host−guest interactions between HKUST-1 and various analytes under different environmental conditions and study the vapor adsorption mechanism by masssensitive and work-function-based readouts. These complementary transduction mechanisms were successfully applied for the detection of low ppm (2 to 50 ppm) concentrations of different alcohols (methanol, ethanol, 1-propanol, and 2-propanol) adsorbed into Cu-BTC thin films. Evaluation of the results allows for the comparison of the amounts of adsorbed vapors and the contribution of each vapor to the changes of the electronic properties of Cu-BTC. The influence of the length of the alcohol chain (C1−C3) and geometry (1-propanol, 2-propanol) as well as their polarity on the sensing performance was investigated, revealing that in dry air, short chain alcohols are more likely adsorbed than long chain alcohols, whereas in humid air, this preference is changed, and the sensitivity toward alcohols is generally decreased. The adsorption mechanism is revealed to differ for dry and humid atmospheres, changing from a site-specific binding of alcohols to the open metal sites under dry conditions to weak physisorption of the analytes dissolved in surface-adsorbed water reservoirs in humid air, with the signal strength being governed by their relative concentration.

M

Gas adsorption and separation capability of this material have already been extensively reported. The influence of polarity and geometry of the target gases on their adsorption behavior on Cu-BTC has been widely discussed, and several adsorption mechanisms have been proposed. According to Karra and Walton,6 Cu-BTC is able to separate molecules with different polarities. Castillo et al.4 and Gutierrez-Sevillano et al.5 simulated adsorption of different gases into the pores of CuBTC and proposed that polar and nonpolar gases preferentially occupy different pores depending on their hydrophilic or hydrophobic nature. Castillo et al.4 proposed different adsorption sites in Cu-BTC, and their simulations show that water molecules preferentially adsorb on the unsaturated metal sites. Simulations of Gutierrez-Sevillano et al.5 show that nonpolar molecules adsorb in the small nonpolar tetrahedral cages, whereas alcohols and water molecules adsorb close to the copper atoms in one of the big polar cages. Van Assche et al.10

etal−organic frameworks (MOFs) span a large group of highly porous, crystalline materials consisting of metal ions that are connected with organic linker molecules to twoor three-dimensional structures.1 Some of them feature open metal sites that might be used for selective gas adsorption based not only on the geometry and size of the molecules but also on their electronic affinity to the metal site.2 Cu-BTC is a very well-investigated MOF in theory and experiment. It consists of Cu ions and benzene tricarboxylate (BTC) as a linker. The crystallographic structure of Cu-BTC was reported by Chui et al.3 Cu-BTC is a three-dimensional framework with a paddle-wheel structure featuring pores with different sizes and shapes. These are large square-shaped pores of 9 Å as well as tetrahedral pores of 5 Å in diameter, which are connected by triangular windows of 3.5 Å.3 Cu-BTC has unsaturated Cu metal sites that can be occupied by water or other polar molecules.4−7 Experimental work of Küsgens et al.,8 who investigated water sorption on Cu-BTC powder, shows that the majority of water molecules can be desorbed from CuBTC under nitrogen flow. The same trend was observed by Biemmi et al.,9 measuring water adsorption with a Cu-BTCcoated quartz crystal microbalance (QCM). © 2014 American Chemical Society

Received: February 26, 2014 Accepted: June 18, 2014 Published: June 18, 2014 6948

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

■

reported about the two-step adsorption of polar adsorbates on Cu-BTC. The majority of the theoretical and experimental work deals with processes that occur under idealized conditions (for example, gas adsorption in dry nitrogen or Cu-BTC storage at elevated temperatures or under nitrogen flow prior to usage). This knowledge provides a basic background for the potential applicability of the materials, but it cannot always be directly transferred to a real application, especially when the material is exposed to ambient conditions for longer periods. Thus, further investigations as a function of the working environment of the sensor are required. For the use of a material as a gas sensor, it is crucial to investigate the gas sensing performance under “realworld” conditions, owing to the fact that at ambient conditions water molecules are present in the atmosphere and are (co)adsorbed on surfaces. The rational choice of suitable transducers is also necessary to guarantee reliable gas detection. Well-known transducers for mass-sensitive detection are QCMs. Biemmi et al.9 grew Cu-BTC thin films directly on the gold electrodes of QCM samples and recorded water sorption isotherms. Khoshaman and Bahreyni11 investigated Cu-BTC for mass-sensitive detection of acetone, tetrahydrofuran and 2-propanol in dry synthetic air. The possibility of selective detection of pentanal with Cu-BTC thick films of ca. 20 μm height using work-function-based readout (Kelvin Probe) was already shown in our group.12 The development and specific modification of work-functionbased readout detection devices like field effect transistors (FETs) for gas sensing applications (GasFETs) has many advantages. GasFETs are not limited to conductive and semiconductive materials as a sensing layer and allow operation at room temperature. Further advantages are compact size and moderate costs during fabrication as well as usage. Many different materials have already been successfully utilized for the application in GasFETs.13,14 The preselection of sensing materials can be done with Kelvin Probe as this transducer is based on the detection of the same physical quantities, i.e. gas adsorption induced changes in the contact potential difference (CPD) between the Kelvin Paddle and the sensing layer. These changes are proportional to the changes in work function ΔΦ. In this work, mass-sensitive readout with QCM as well as work-function-based detection with Kelvin Probe were combined for the evaluation of the adsorption behavior of Cu-BTC toward low ppm concentrations of different alcohol vapors (methanol, ethanol, 1-propanol, and 2-propanol). QCM records changes in resonant frequency caused by the adsorption of molecules in the sensing layer independent of their interactions, thus allowing for the quantification of the amount adsorbed. The work-function-based detection with a Kelvin Probe setup “visualizes” the changes in electronic properties (work function, ΔΦ) of the sensing layer caused by gas adsorption and can depend on the type of adsorption (weak physisorption often causes minor changes in ΔΦ compared to strong chemisorption). The adsorption mechanism of these gases on Cu-BTC as well as the influence of the length of the alcohol chain (C1−C3) and geometry (1-propanol, 2propanol) were investigated in dry and in humid synthetic air at room temperature. The combination of both readout methods is helpful for a better understanding of the adsorption mechanism in order to verify the applicability of Cu-BTC for selective gas detection.

Article

MATERIALS AND METHODS

PREPARATION OF Cu-BTC NANOPARTICLES. Cu-BTC particles were synthesized in a fashion similar to a literature protocol.15 H3BTC (0.738 g, 3.44 mmol) was dissolved in ethanol (14 mL) and DMF (42 mL) and combined with PAA (2.21 g, 1.23 mmol). To this mixture, a solution of Cu(OAc)2 •H2O (0.7 g, 3.44 mmol) in 28 mL of deionized water was added under vigorous stirring, which rapidly induced the formation of a blue precipitate. After 30 min, the product was separated from the reaction mixture by centrifugation and washed in DMF for at least three times. Stable colloidal suspensions of Cu-BTC were obtained by redispersing the washed product in DMF using ultrasound. PREPARATION OF Cu-BTC SENSING LAYERS. Al2O3 substrates with thin films of sputtered TiN serving as a back electrode were used for the preparation of Kelvin Probe samples. Vibrating quartzes with a nominal resonant frequency of 10 MHz featuring gold as a back electrode were used for the preparation of the QCM samples. Cu-BTC sensing layers were drop-coated on the back electrodes of Kelvin Probe and QCM samples, respectively, and subsequently dried. Two microliters of the synthesized CuBTC suspension were used for the preparation of each sample. The prepared samples were stored at ambient conditions and used for the evaluation of gas sensing without further treatment. To adapt the samples to the measurement conditions (dry or humid synthetic air), the samples were exposed to synthetic air (dry or humid, depending on the subsequent measurement) in the first segment of the measurement cycle for several (1−5) hours. CHARACTERIZATION METHODS. Dried Cu-BTC samples were characterized by powder X-ray diffraction (XRD), infrared (IR) spectroscopy, and scanning electron microscopy (SEM). XRD measurements were carried out on a Huber G670 diffractometer in Guinier geometry using either Cu Kα1 (λ = 1.54051 Å) or Co Kα1 radiation (λ = 1.788965 Å; both Ge(111)-monochromated). IR spectra were collected with a PerkinElmer Spektrum BX II spectrometer with an attenuated total reflectance unit. SE micrographs were recorded with a JEOL JSM-6500F SEM. KELVIN PROBE AND QCM SETUP. QCM measurements were performed with JLMQ USB Interface for QCM sensors (JLM Innovation, Germany). According to Sauerbrey,16 a QCM can be coated with a sensitive layer and used for masssensitive detection of different analytes by measuring changes in the oscillation frequency. These frequency changes (Δf) are proportional to the mass loading (Δm) resulting from adsorption or desorption of analytes if the changes in the resonance frequency between coated and uncoated QCM samples are below 2%. m = −c*Δf

(1)

(c is a constant that summarizes geometrical and physical constants of the quartz crystals used). Kelvin measurements were performed using a Kelvin Probe setup (Besocke Delta Phi, Germany). The measured signal represents the CPD between the oscillating gold paddle and the sensing layer. This CPD is proportional to the changes in work function: CPD = 6949

1 ΔΦ e

(2) dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

Figure 1. Experimental setup for the gas adsorption measurements in flow mode.

(e is the unit charge). The work function (Φ) is an energy difference between the Fermi (EF) and the vacuum level (EVAC). In a semiconductive or insulating material, the Fermi level is not occupied by electrons, and thus further energy terms are introduced for the correct description of the work function: Φ = E VAC − E F = (EC − E F)b + χ − eΔVs

n=

m Mm

(4)

For this reason, some evaluations of changes in the resonant frequency (Δf) are normalized to the molecular weight (Mm) of each alcohol, respectively (Δf/Mm), in the case where the comparison of the adsorbed amount of molecules is of interest.

■

(3)

RESULTS AND DISCUSSION CHARACTERIZATION OF Cu-BTC SENSING LAYERS. An as-prepared Cu-BTC suspension was analyzed by XRD before

(EC is conduction band level, χ is the electron affinity and eΔV is the band bending. “b” indicates bulk and “s” surface.)17 Adsorption of the molecules leads to formation of additional surface states and/or dipoles and consequently contributes to changes in the work function by changing one or more of its terms. Kelvin Probe represents changes in work function between the Kelvin paddle and the sensing layer. If the reaction of the gold paddle and the back electrode with the gas is negligible, it can be concluded that the measured signal represents the changes of the electronic structure of the sensitive layer under investigation. Target gases (a specific concentration of a target gas diluted in nitrogen) were purchased by LindeGas (Pullach, Germany). All measurements were performed in synthetic air (80% nitrogen, 20% oxygen) with varying relative humidity (r.h.) and a total gas flow of 1 L/min. The desired level of relative humidity was achieved by directing a part of the nitrogen flow through the bubbler filled with deionized water. The humidity sensor (Sensirion, Switzerland) was used for the monitoring of the humidity level (Figure 1). Technical parameters of the mass flow controller and the target gas concentration in the gas bottle determine the possible adjustment of the minimal and maximal target gas concentration. The reliability of the gas transport through the setup and the determination of the real target gas concentration was proved with portable Fourier transformed Infrared analyzer (Ansyco DX4015, Germany), which was inserted after the measurement chamber. DATA PROCESSING. The QCM response represents the changes in the resonant frequency (Δf) caused by the mass loading (Δm). For the direct comparison of the amount of molecules adsorbed on the sensing layer, the molecular weight of the molecules should be considered. The amount of the molecules n can be calculated if their mass m and molecular weight Mm are known (eq 4).

Figure 2. XRD pattern of Cu-BTC. Dried Cu-BTC sample (red line) and the simulated Cu-BTC (black line).

sensor coating. The XRD analysis (Figure 2) of the dried solid confirms the identity of the synthesized material, as well as the nanoscale size of the particles which we deduce from the notable peak broadening. The differences in the reflection intensities most likely result from differences in the activation state of the sample;18,19 the pores are partially filled with solvent (here DMF) in the measured sample, which was not specifically activated prior to the XRD measurement and stored under ambient conditions. The water stability of the particles at ambient conditions was tested by exposing dried Cu-BTC samples to humid air and comparing the analytical data with a reference sample stored in a closed vial for 5 months (see Supporting Information for XRD patterns, IR spectra and SEM images, Figures SI 1 and SI 2, respectively). Figure 3 depicts optical images of Cu-BTC-coated QCM (Figure 3 a) and Kelvin (Figure 3 b) samples. SEM images of a Cu-BTC layer on QCM and Kelvin samples are shown in Figure 4. As prepared Cu-BTC films on the gold electrodes of QCM samples exhibit domain formation. Their layer thickness varies 6950

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

Figure 3. (a) Optical images of a Cu-BTC-coated QCM sample and (b) of a Cu-BTC-coated Kelvin sample.

Figure 4. SEM images. (a, b) Surface and cross section of Cu-BTC thin films on the gold QCM electrode, (c, d) Surface and cross section of CuBTC thin films on the TiN electrode of the Kelvin Probe sample.

between 1.0 and 1.7 μm. The Cu-BTC layers on the TiN electrode are homogeneous, and their layer thickness varies between 1.2−2.5 μm. The particle size is in the range between 50 and 80 nm. The resonant frequencies of uncoated and Cu-BTC-coated QCM samples were measured at ambient conditions. The changes in resonance frequency of Cu-BTC-coated QCM samples were ascertained to be below 2%, indicating the applicability of the Sauerbrey equation.16

To verify that the Kelvin Probe signal represents the response of the Cu-BTC sensing layer only, rather than being caused by the back electrode, uncoated TiN Kelvin Probe samples were measured first. Exposure of the uncoated TiN back electrode to the target gases used in this study show no changes in work function. Hence, it can be concluded that the changes in work function recorded with the Kelvin Probe setup are caused by the sensing layer. 6951

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

after exposure of Cu-BTC to liquid water for 24 h, indicating degradation of the material. Aging-induced changes in sensitivity of Cu-BTC are generally not well-investigated yet. According to Küsgens et al.,8 the water stability of Cu-BTC might be sufficient for molecular recognition if the framework is not directly exposed to liquid water. This would be sufficient for the application of Cu-BTC as a sensing layer at ambient conditions. Moreover, the recent work of Majano et al.23 demonstrates the possibility of reconstruction of Cu-BTC by exposing the degraded material to a stream of pure ethanol or ethanol/water mixtures. We investigated the water stability of our particles at ambient conditions by exposing dried Cu-BTC samples to humid air and comparing the analytical data with a reference sample stored in a closed vial for 5 months. The XRD patterns and IR spectra measured after 10 days of exposure (see Figure SI 1, Supporting Information) reveal no pronounced degradation of the structure except for an increase of the IR band around 3500 cm−1, indicating enhanced water uptake. SEM images of the solid samples and a QCM sample exposed to different target gases over 9 months (Figure SI 2 (a−c), Supporting Information) show no obvious changes in the appearance of the material in comparison to the sample stored in the vial (Figure SI 2 d), Supporting Information) and a “fresh” QCM sensor (see Figure 4 a, b), respectively. Although the contact with a humidity level similar or even higher than at ambient conditions has shown no severe effect on the sample crystallinity for the investigated period, a potential aging effect, however, due to the prolonged storage at ambient humidity as well as the exposure to different target gases cannot be excluded. The measurements presented in the current study were performed over a period of several months. Here, aging of the Cu-BTC framework was also observed during this period of time. The signal heights in both mass-sensitive and workfunction-based responses decreased with increasing time delay between sample preparation and measurement (see Supporting Information, Figure SI 3, 4). However, it should be noted that the relative order of the signal intensities during exposure to different alcohol vapors was unchanged (see Supporting Information, Figure SI 3, 4). To eliminate the influence of the possible alteration of the samples during that time, the quantitative evaluation of the results was done within one measurement or by comparing, back to back, several subsequent measurements recorded in shorter periods (1−3 weeks). Due to the limited possibility to connect more than

Figure 5. First three test cycles of the Kelvin and QCM measurements with methanol, ethanol, and 2-propanol as target gases at 40% r.h.

The area of the QCM sample is similar to the area of the Kelvin Probe sample. As the same amount of the Cu-BTC suspension was used for the preparation of both samples and consequently should have a similar amount and accessibility of adsorption sites, the results obtained with both methods can be compared. REPRODUCIBILITY OF QCM AND KELVIN PROBE MEASUREMENTS. Several studies report on the degradation and structural alteration of the Cu-BTC framework. Van Assche et al.10 observed that the gravimetric ethanol uptake of Cu-BTC (measured at 323 K) is lower for a sample that was exposed to the atmosphere longer than 10 months than for a sample that was stored in the activated state. Pöppl et al.20 observed a loss of crystallinity and porosity of Cu-BTC after contact to moist air for a period of 49 days. The loss of 26% of the surface area of Cu-BTC was observed by Schoenecker et al.,21 who compared BET surface areas of Cu-BTC before the exposure to 80% of relative humidity and after measurement of sorption isotherms and reactivation. Irreversible changes of Cu-BTC after its immersion in liquid water at 323 K for 24 h were also observed in XRD investigations of Küsgens et al.8 Similar behavior of Cu-BTC toward water is reported by Cychosz and Matzger.22 The authors of ref 22 compared the XRD patterns of Cu-BTC before and after exposure to water and observed no changes after exposure of Cu-BTC to humid air (18% r.h.) for 24 h, whereas additional peaks in the powder XRD appeared

Figure 6. Mean value and standard deviations for the exposure to 30 ppm of each gas (a) Kelvin 40% r.h., (b) QCM 40% r.h. 6952

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

Figure 7. Work function (ΔΦ) response (a) and normalized QCM response (Δf/Mm) (b) to 50 ppm of methanol, ethanol, 1-propanol, and 2propanol, respectively.

Figure 8. Characteristic response curves. (a) Kelvin measurement at r.h. 40% (detectable changes in the work function in response to methanol could be observed at concentrations only above 25 ppm). (b) Normalized QCM response at r.h. 40%. (c) Kelvin measurement at 0% r.h. (d) Normalized QCM response at r.h. 0%. Connecting lines between the points serve for clearness.

To eliminate the possible influence of the sequence of the gas exposure as well as to verify the reproducibility and reliability of the recorded data, the samples were exposed to 2-propanol followed by methanol and ethanol and again by 2-propanol and so on for both Kelvin Probe and QCM measurements. These cycles were repeated six times at 40% r.h., and the height of the Kelvin Probe and QCM signal was evaluated. Figure 5 displays the first three cycles of this measurement. Figure 6 shows the

three test gases to the gas measurement setup simultaneously, two different combinations of target gases were selected. In the first run methanol, ethanol and 2-propanol were used and compared, whereas in the second run ethanol, 1-propanol and 2-propanol were used. The same sequences of the signal intensities were obtained with different combinations of target gases indicating the retention of the relative signal intensities. 6953

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

Figure 9. Linear fit of (a) Kelvin measurement at 40% r.h. and (b) QCM measurement at 40% r.h.

Figure 10. Linear fits of (a) Kelvin measurement at 0% r.h. and (b) QCM measurement at 0% r.h.

Figure 11. (a) Kelvin (red line) and QCM (black line) measurements. ΔΦ and Δf responses to 10 and 50 ppm of (a) methanol, ethanol, and 2propanol and (b) ethanol, 1-propanol, and 2-propanol, respectively, at 40% r.h.

mean value and the standard deviation for the exposure to 30 ppm of each gas. The negative slope of the curves indicates that the equilibrium (after placing the sample into the measurement chamber) is not completely reached. Nevertheless, the onset of the analyte sorption events is clearly discernible from the baseline. It can be observed for both QCM and Kelvin Probe measurements that the standard deviation is smaller than the changes in the signal height between different gases.

Our results show that differences in work function (ΔΦ) and resonant frequencies (Δf) responses to different alcohols are independent of the sequence of exposure to these vapors. The differences between the mean values of ΔΦ and Δf are significantly larger than their standard deviations, which allows us to distinguish these vapors on the basis of their signal height for further evaluations. 6954

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

Figure 12. (a) Kelvin (red line) and QCM (black line) measurements. ΔΦ and Δf responses to 10 and 50 ppm (a) of methanol, ethanol and 2propanol and (b) ethanol, 1-propanol, and 2-propanol, respectively, at 0% r.h.

2-propanol are presented in Figure 7. The connection lines between the points serve as guides to the eye. Work function (ΔΦ) response as well as the normalized changes in resonant frequency (Δf/Mm) decrease with increasing level of humidity for all alcohols investigated. Kelvin measurements show the highest signal for 1-propanol, followed by 2-propanol and ethanol for all levels of humidity. The Δf/Mm response is different for humidity levels below and above a r.h. of 20%. At lower humidity levels, the adsorption of methanol and ethanol is prevailing, at higher humidity levels 1propanol and 2-propanol adsorb more strongly and the difference between the amounts of the adsorbed vapors becomes smaller. CHARACTERISTIC RESPONSE CURVES. The majority of vapor adsorption experiments and simulations reported in the literature were performed in dry atmosphere. To be able to compare our results with results reported in the literature as well as to investigate the influence of humidity on alcohol adsorption, our measurements were performed in both dry and humid synthetic air. To estimate the dependency of the signal height on vapor adsorption, mass-sensitive and work function responses for different (2 to 50 ppm) concentrations of methanol, ethanol, 1-

Table 1. Sensitivity and Response to 10 ppm of Each Alcohol at 40% r.h. sensitivity and response at 40% r.h.

methanol

ethanol

1-propanol

2-propanol

Kelvin mV/ppm value for 10 ppm QCM Hz/ppm value for 10 ppm

0.02 1 (for 40 ppm) 0.05 1.2

0.07 1.3 0.2 4.0

0.17 2.9 0.3 7.6

0.13 1.9 0.2 3.4

Table 2. Sensitivity and Response to 10 ppm of Each Alcohol at 0% r.h. sensitivity and response at 0% r.h.

methanol

ethanol

1-propanol

2-propanol

Kelvin mV/log ppm value for 10 ppm QCM Hz/log ppm value for 10 ppm

10.1 9.3 158.0 226.5

13.2 14.4 140.6 196.1

14.3 21.8 57.8 79.5

13.2 15.5 37.6 38.4

INFLUENCE OF HUMIDITY ON ALCOHOL SENSING. As a next step, the influence of the degree of the relative humidity on alcohol sensing was investigated. The characteristic curves for exposure to 50 ppm of methanol, ethanol, 1-propanol, and

Figure 13. Simultaneous evaluation of Kelvin and QCM responses (a) at 40% r.h. and (b) at 0% r.h. 6955

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

pressures, followed by ethanol, 1-propanol, and 2-propanol at 0% r.h. The two-step adsorption described in ref 10 is not expected in our measurement due to the low analyte concentrations used. The decreasing number of adsorbed alcohol molecules with increasing chain length, mass, and size is also in agreement with ref 21, which reports about the affinity of Cu-BTC to small molecules with accessible lone pairs of electrons. Gutiérrez-Sevillano et al.5 simulated preferential adsorption sites in the Cu-BTC MOF for different molecules including methanol, ethanol, propanol, and water. According to their simulations, the majority of these molecules adsorb on the unsaturated metal sites in the largest pores of Cu-BTC, but other pores and adsorption sites are also occupied. In contrast to QCM measurements, the work function response (Figure 10a) increases with increasing length of the carbon chain of the respective alcohol and is higher for 1propanol than for 2-propanol. Although less 1-propanol and 2propanol are adsorbed on Cu-BTC according to our QCM measurements (Figure 10b), their contribution to the changes in electronic properties is apparently more pronounced. One possible explanation for these observations may be the different adsorption behavior for the larger and smaller analytes, respectively. According to ref 24, the lengths of methanol and ethanol molecules are 3.3 and 4.6 Å, respectively, and their diameters are 2.8 and 3.6 Å, respectively. These geometries will allow for the adsorption of these molecules into the small pores of 5 Å. Hence, although all molecules adsorb into the big pores where they strongly interact with unsaturated Cu metal sites, giving rise to pronounced work function changes, the small molecules can additionally adsorb more easily into the small and big pores that are already occupied by analyte without showing a strong interaction with the host lattice (weak physisorption). This incorporation will have the same contribution to the mass changes, whereas the contribution to the changes in work function may be reduced. As a consequence, the same amount of adsorbed small molecules as compared to the larger ones may not give rise to as pronounced changes in the work function. According to Van Assche et al.10, the number of adsorbate molecules per unit cell of Cu-BTC based on the maximum vapor phase uptake at 323 K decreases from methanol to ethanol to 1-propanol to 2-propanol and is also in agreement with our QCM measurements. The positive inductive effect of alcohols25 and increasing adsorption enthalpies with increasing length of the carbon chain5 (the calculated adsorption enthalpy values at 295 K according to ref 5 are 42.6 kJ/mol for methanol, 45.7 kJ/mol for ethanol, and 52. 6 kJ/mol for propanol) as well as the additional van der Waals interaction of alkyl chains with the pore surface of the MOF, as proposed for different hydrocarbon interactions with MIL-47 MOF26, may also contribute to the observed larger changes in work function for longer alcohols. The differences in the amount of adsorbed 1-propanol and 2-propanol molecules can be explained by the geometry of these molecules in relation to the geometry of the pores and their polarity. For example, 1propanol is an elongated molecule and can more easily penetrate into the pores than 2-propanol, which is more bulky compared to 1-propanol. The slightly higher polarity of 1-propanol as compared to 2-propanol may further assist the more facile uptake of the former into both the smaller and more “polar” larger pores of HKUST-1. In humid air, both QCM and Kelvin Probe responses increase linearly with increasing alcohol concentration (Figure 9a,b). The total signal height is much lower in comparison to

propanol, and 2-propanol were measured in dry and humid (r.h. 40%) synthetic air. The signal height after 10 min of vapor exposure was evaluated. The time that is needed to reach the equilibrium (a plateau in the time dependent response) depends on the type and concentration of the target vapors as well as on the measurement conditions (dry or humid). Not all alcohols used are equilibrated after 10 min, but major changes were consistently achieved within several minutes (see Supporting Information, Figure SI 5). As the current study focuses on the possibility of integration of the Cu-BTC sensing layer in chemical sensors, the evaluation of the responses to target gases within a short period of time is of interest. The obtained characteristic curves represent the mean value and the standard deviation calculated from two separate measurements for methanol, ethanol, 1-propanol, and 2-propanol (with the exception of ethanol in dry air, which was measured three times) and are presented in Figure 8. Characteristic curves of the work function (ΔΦ) (Figure 8 a,c) and normalized QCM responses (Δf/Mm) (Figure 8 b,d) show an increase with increasing concentration of alcohols. Major differences can be observed between dry and humid air. Measurements with both methods show higher response in dry air in comparison to the response in humid air. It can be observed that the characteristic curves for all alcohols show qualitatively similar shapes for dry and humid synthetic air, respectively. Moreover, the shapes of the characteristic curves for Kelvin Probe and QCM measurements are also similar. The ΔΦ response increases with increasing length of the carbon chain of alcohol for each concentration in both humid and dry air. In addition, the ΔΦ response to 1-propanol is higher than for 2-propanol (Figure 8 a,c). The normalized QCM response (Δf/Mm) shows differences in dry and humid synthetic air regarding the amount and preference for the adsorbed molecules. In humid air, 1-propanol is predominantly adsorbed on the Cu-BTC, followed by ethanol and 2-propanol, and methanol. In dry air, in contrast, the amount of adsorbed alcohol molecules decreases with increasing length of the alcohol chain and is higher for 1-propanol than for 2-propanol. The characteristic curves measured in humid air show approximately a linear increase in ΔΦ as well as in Δf/Mm response with increasing gas concentration. This linear increase corresponds to the Henry isotherm. The linear fit is presented in Figure 9 a,b. Characteristic curves in dry air show a linear behavior in a semilogarithmic plot (Figure 10 a,b). Both ΔΦ and Δf/Mm responses increase linearly with a decade of concentration. The linearity of the semilogarithmic plot corresponds to the Temkin isotherm, indicating decreasing adsorption enthalpies with increasing surface coverage. Similar slopes of the curves as in Figure 10 b were also observed by Van Assche et al.10 for low vapor pressures of alcohols. Van Assche et al.10 measured gravimetric uptake of vapors of different alcohols on Cu-BTC in dry nitrogen at 323 K. The highest uptake capacity (measured in milligram of adsorbed amount per gram of Cu-BTC) for low vapor pressures (= low vapor concentrations) was observed for 1hexanol, followed by 1-butanol, 1-propanol, 2-propanol, ethanol, and methanol, and the authors report that less polar alcohols have higher uptake at lower pressure. We would like to note that the authors compare the mass of the adsorbate, rather than the number of adsorbed molecules. By normalizing these results by the molecular weight of each molecule, the same trend as in our QCM measurements is expected. Here, more methanol molecules are adsorbed in Cu-BTC at lower 6956

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

simultaneously are influenced only to a minor extent by the presence of another alcohol species in the atmosphere and are nearly the sum of the signals during individual exposure to each alcohol (see Supporting Information, Figure 6). In dry atmosphere, in contrast (Figure 12), ΔΦ as well as Δf responses are strongly influenced by the presence of another alcohol species in the atmosphere. The total response to the simultaneous exposure to two alcohols is smaller than the sum of their individual responses. The calculated values can be obtained from Supporting Information (Figure SI 7). The results obtained with Kelvin Probe and QCM measurements where Cu-BTC was exposed to different alcohols simultaneously are in agreement with our assumption that in dry air the alcohols are predominantly adsorbed on the unsaturated metal sites of Cu-BTC, although in humid air, they are dissolved in water that is adsorbed on the surface and inside the pores and should be regarded as an alcohol−water mixture. QCM as well as Kelvin Probe measurements performed in humid air show that the adsorption of an alcohol is independent of the presence of a second alcohol in the atmosphere, and consequently, the amount of the alcohol dissolved in water is only dependent on its partial pressure (= concentration) in the gas phase. In dry air, the first alcohol occupies specific adsorption sites in Cu-BTC, and consequently, fewer adsorption sites are available for the adsorption of the second alcohol. This results in the observed smaller signal upon exposure to the second alcohol with both Kelvin as well as QCM readout. The comparison of the shape of the QCM and Kelvin signals in dry and humid air confirm the proposed adsorption mechanism. SENSITIVITY. The sensitivity defined as change in work function ΔΦ per change in concentration (slope of the linear curve) can be determined from Figure 10 a for Kelvin Probe measurements. For the definition of the sensitivity in a QCM measurement, the non-normalized Δf response was used (Figure 10 b). Tables 1 and 2 summarize the sensitivity toward different alcohols for Kelvin Probe and QCM measurements, as well as the value of the ΔΦ and Δf responses for 10 ppm of each gas in humid and dry air. The highest sensitivity at 40% r.h. is observed for 1-propanol with both methods. In dry synthetic air, Kelvin Probe measurements show the highest sensitivity to 1-propanol, although with QCM, the highest sensitivity is observed for methanol. SELECTIVITY. Figure 13 shows the simultaneous evaluation of the readout of both methods for humid (Figure 13 a) and dry air (Figure 13 b). The single points represent the simultaneously recorded responses of Kelvin Probe and QCM measurements for different concentrations (2 to 50 ppm) of alcohol vapors. It can be observed that the simultaneous evaluation of the QCM and Kelvin Probe response in dry air allows for the clear discrimination between alcohols with shorter and larger alkyl chains. In humid air, the discrimination is only possible for larger values that according to Figure 8 correspond to higher alcohol concentrations.

dry air. The work function response again increases with increasing length of the alcohol chain and is higher for 1propanol than for 2-propanol. The linear dependency of gas adsorption on the partial pressure of a gas is described by a Henry isotherm.27 In contrast to dry air, the evaluation of the QCM response in humid air shows that 1-propanol is preferably adsorbed, followed by ethanol and 2-propanol, and methanol, indicating different adsorption mechanisms in dry and humid air, which is rationalized as follows. In synthetic air with sufficient relative humidity, the water molecules are not only adsorbed on the unsaturated metal sites of Cu-BTC4 but also form hydrogen bonds between each other that can result in formation of water clusters.8 Additionally, water molecules can be adsorbed in textural pores between the Cu-BTC grains as well as on the surface. All investigated alcohols are completely soluble in water.28 According to Henry,29 the solubility of gas in the liquid is directly proportional to its partial pressure above the liquid. If the pressure above the liquid phase becomes smaller, the gas molecules will escape from the liquid phase to equilibrate with the gas phase. Alcohols can form hydrogen bonds to water molecules on the film surface, on the internal surfaces and textural pores, and to the water molecules inside the pores of Cu-BTC. The linear dependency of the QCM and Kelvin Probe signals obtained in humid air therefore indicates the solubility of alcohols in water, and the uptake behavior in humid air is dominated by rather unspecific dissolution of the alcohols in the different types of water reservoirs. Moreover, the vapor pressure of 1-propanol is lower than that of 2-propanol, which might result in its preferential condensation on the sensing layers. However, beside concentration effects, the complex interplay between geometrical and chemical factors such as size and polarity of the pores and molecules, respectively, as well as nonidealities of the binary water− analyte mixtures will additionally contribute to the signal height obtained. For instance, as the adsorption enthalpy of water on Cu-BTC is also comparatively high,5 a preoccupation of potential adsorption sites at 40% r.h. can be expected. The signal intensity of a certain sorptive will then be controlled by its individual properties (such as steric effects and interaction strength) in comparison to those of water, both competing for the same binding sites. This assumption is also in agreement with Figure 7 b, which shows the strong influence of the humidity level above 20% on the adsorbed amount of different alcohols. SIMULTANEOUS EXPOSURE TO DIFFERENT ALCOHOLS. The following measurements were performed to prove the proposed adsorption mechanism. Different alcohols were simultaneously and successively added to synthetic air to investigate the influence of the preoccupation of adsorption sites by one alcohol on the adsorption of the other alcohol. Two different test protocols were used. In the first run, one alcohol was added to synthetic air, and after 10 min of exposure, a second alcohol with the same concentration was added. The flow of the second alcohol was switched off after 10 min of exposure, whereas the flow of the first alcohol was still “on”. Then, 10 or 20 min later, the flow of the third alcohol was switched on for 10 min and so on. This procedure was repeated for all alcohols. In the second run, two alcohols with the same concentration were switched on simultaneously for 10 min. The results in humid synthetic air (Figure 11) show that changes in ΔΦ as well as in Δf during exposure to two alcohols

■

CONCLUSION AND OUTLOOK Cu-BTC thin films were investigated as gas sensing layers for the detection of low (2 to 50 ppm) concentrations of different alcohols (methanol, ethanol, 1-propanol, and 2-propanol) at room temperature with mass-sensitive (QCM) and workfunction-based (Kelvin Probe) readouts. Our results show that alcohols can be successfully detected with both methods at ppm 6957

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958

Analytical Chemistry

Article

(12) Davydovskaya, P.; Pohle, R.; Tawil, A.; Fleischer, M. Sens. Actuators, B 2013, 187, 142−146. (13) Fleischer, M. Meas. Sci. Technol. 2008, 19, 042001. (14) Oprea, A.; Bârsan, N.; Weimar, U. Sens. Actuators, B 2009, 142, 470−493. (15) Ranft, A.; Betzler, S. B.; Haase, F.; Lotsch, B. V. CrystEngComm 2013, 15, 9296−9300. (16) Sauerbrey, G. Z. Physik 1959, 155, 206−222. (17) Henzler, M.; Göpel, W. Oberflächenphysik des Festkörpers; Vieweg+Teubner Verlag: Wiesbaden, 1994. (18) Borfecchia, E.; Maurelli, S.; Gianolio, D.; Groppo, E.; Chiesa, M.; Bonino, F.; Lamberti, C. J. Phys. Chem. C 2012, 116, 19839− 19850. (19) Hafizovic, J.; Bjørgen, M.; Olsbye, U.; Dietzel, P. D. C.; Bordiga, S.; Prestipino, C.; Lamberti, C.; Lillerud, K. P. J. Am. Chem. Soc. 2007, 129, 3612−3620. (20) Pöppl, A.; Jee, B.; Icker, M.; Hartmann, M.; Himsl, D. Chemie Ingenieur Technik 2010, 82, 1025−1029. (21) Schoenecker, P. M.; Carson, C. G.; Jasuja, H.; Flemming, C. J. J.; Walton, K. S. Ind. Eng. Chem. Res. 2012, 51, 6513−6519. (22) Cychosz, K. A.; Matzger, A. J. Langmuir 2010, 26, 17198− 17202. (23) Majano, G.; Martin, O.; Hammes, M.; Smeets, S.; Baerlocher, C.; Pérez-Ramírez, J. Adv. Funct. Mater. 2014, DOI: 10.1002/ adfm.201303678. (24) Wu, H.; Gong, Q.; Olson, D. H.; Li, J. Chem. Rev. 2012, 112, 836−868. (25) Davydovskaya, P.; Pentyala, V.; Yurchenko, O.; Hussein, L.; Pohle, R.; Urban, G. A. Sens. Actuators, B 2014, 193, 911−917. (26) Rosenbach, N., Jr; Ghoufi, A.; Deroche, I.; Llewellyn, P. L.; Devic, T.; Bourrelly, S.; Serre, C.; Ferey, G.; Maurin, G. Phys. Chem. Chem. Phys. 2010, 12, 6428−6437. (27) Butt, H.-J.; Graf, K.; Kappl, M. In Physics and Chemistry of Interfaces; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, 2004; pp 177−205. (28) Yaws, C. L.; Hopper, J. R.; Sheth, S. D.; Han, M.; Pike, R. W. Waste Management 1997, 17, 541−547. (29) Henry, W. Philos. Trans. R. Soc. London 1803, 93, 29−274.

concentration levels. Characteristic curve for each alcohol were recorded, and their evaluation indicates different alcohol adsorption mechanisms in dry and humid air. It was shown that in dry air, alcohols with longer carbon chains are less frequently adsorbed on Cu-BTC, although the work function response increases with increasing length of the alcohol chain. The semilogarithmic characteristic curve recorded in dry synthetic air with QCM and Kelvin measurements indicate alcohol adsorption on the unsaturated Cu metal sites. Smaller changes in work function to more frequently adsorbed methanol and ethanol molecules might be caused by their additional physisorption into the small and big pores without showing a strong interaction with the host lattice. The linear characteristic curve in humid air and the increasing amount of adsorbed alcohols with increasing carbon chain length indicates dissolution of alcohols in the adsorbed water. The next step will be the implementation of Cu-BTC thin films in commercial work-function-based readout transducer devices like gas-field effect transistors (GasFETs).

■

ASSOCIATED CONTENT

S Supporting Information *

Additional information as noted in the text. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected], polina. [email protected]. Fax: +49(89)63646881. Tel.: +49(89)63641603. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The financial support from the German Federal Ministry of Education and Research in the course of the project NanoGasFET (16SV5378K) is acknowledged. We thank the Deutsche Forschungsgemeinschaft for funding within the priority program “Metal−organic frameworks” (SPP-1362).

■

REFERENCES

(1) Metal-Organic Frameworks; MacGillivray, L. R.; John Wiley & Sons, Inc.: Hoboken, NJ, 2010. (2) Yaghi, O. M.; Li, H.; Groy, T. L. J. Am. Chem. Soc. 1996, 118, 9096−9101. (3) Chui, S. S.-Y.; Lo, S. M.-F.; Charmant, J. P. H.; Orpen, A. G.; Williams, I. D. Science 1999, 283, 1148−1150. (4) Castillo, J. M.; Vlugt, T. J. H.; Calero, S. a. J. Phys. Chem. C 2008, 112, 15934−15939. (5) Gutiérrez-Sevillano, J. J.; Vicent-Luna, J. M.; Dubbeldam, D.; Calero, S. J. Phys. Chem. C 2013, 117, 11357−11366. (6) Karra, J. R.; Walton, K. S. Langmuir 2008, 24, 8620−8626. (7) Terencio, T.; Di Renzo, F.; Berthomieu, D.; Trens, P. J. Phys. Chem. C 2013, 117, 26156−26165. (8) Küsgens, P.; Rose, M.; Senkovska, I.; Fröde, H.; Henschel, A.; Siegle, S.; Kaskel, S. Microporous Mesoporous Mater. 2009, 120, 325− 330. (9) Biemmi, E.; Darga, A.; Stock, N.; Bein, T. Microporous Mesoporous Mater. 2008, 114, 380−386. (10) Van Assche, T. R. C.; Duerinck, T.; Gutiérrez Sevillano, J. J.; Calero, S.; Baron, G. V.; Denayer, J. F. M. J. Phys. Chem. C 2013, 117, 18100−18111. (11) Khoshaman, A. H.; Bahreyni, B. Sens. Actuators, B 2012, 162, 114−119. 6958

dx.doi.org/10.1021/ac500759n | Anal. Chem. 2014, 86, 6948−6958