- Email: [email protected]
Time Dependence of Vibrational Relaxation of Deuterated Hydroxyls in Acidic Zeolites.
J. Weitkamp, H.G. Karge, H. Pfeifer and W. Hdlderich (Eds.)
Zeolites and Related Microporous Maierials: Siaie of ihe Ari 1994 Studies in Surface Science and Catalysis, Vol. 84 0 1994 Elsevier Scicnce B.V. All rights reservcd.
493
Time Dependence of Vibrational Relaxation of Deuterated Hydroxyls in Acidic Zeolites. Mischa Bonn 1, Marco J.P. Brugmans', Aart W. Kleyn', Rutger A. van Santent and Ad Lagendijk* *I
* FOM-Institute for Atomic and Molecular Physics, Kruislaan 407, 1098 SJ Amsterdam,
Tbe Netherlands t Scbuit Institute of Catalysis, Eindboven University of Tecbnology, P.O. Box 513, 5600 MB Eindboven, The Netberlands Abstract. In this picosecond time-resolved study of zeolite hydroxyls we conclude from the 0-D vibrational population lifetime that the protons in Y zeolite as well as in Mordenite are hydrogen bonded. In deuterated ZSM-5no hydrogen-bonding was observed.
Introduction. Acidic zeolites are widely applied in hydrocarbon conversion reactions in the petrochemical industry. Their acidity is due to Brensted active hydroxyl groups, which are the catalytically active sites. The hydroxyl group is located between a silicon and an aluminum atom. In order to gain deeper insight into the dynamic properties of these sites, we have investigated the relaxation of vibrationally excited deuterated hydroxyl groups with time-resolved (picosecond) infrared saturation spectroscopy. Fkom our experiments we obtain 2'1, the vibrational lifetime, which cannot be obtained by conventional infrared spectroscopy, due to the fact that the width of the absorption bands is determined by inhomogeneous broadening. The population lifetimes are in the order of 100 ps [l-41, corresponding to a bandwidth of 0.05 cm-', whereas the absorption linewidths are in the order of 30 cm-'. The broadening is not due to dephasing; the behaviour of the width of the absorption peak with temperature can by no means be described by a theory of 2'2 broadening [5]). The relation between acidity, absorption frequency and local environment of the zeolite hydroxyl has been the subject of ongoing research. Lately, a great deal of attention has been focused on this question in zeolite Mordenite.[6-111 We investigated the wavenumber dependence of the 0-D vibrational population lifetime 21' for several zeolites, in order to obtain new information concerning the nature of the inhomogeneous broadening of the 0-D absorption peaks. Calculations were performed to support our findings. Experimental Section. In our pump-probe experiments a considerable fraction ( ~ 1 0 - 2 0 % )of the 0-D oscillators is excited from their v = 0 to 'u = 1vibrational state by a powerful short laser pulse. As a result the 0-D absorption is bleached on picosecond time scales; the transmission of light
494
of this wavelength through the sample increases temporarily. Hence the equilibration of the population-distribution can be monitored by measuring the transmission of a much weaker pulse -the probe pulse- whose time delay with respect to the pump-pulse can be varied. The decay of the pump-induced transmission with time is related to the vibrational lifetime of the excitation 21' , as: ln[T(t)/To] exp(-t/Tl), where T ( t )is the transmitted energy of the probe pulse at delay t and 0'2 is the transmitted probe energy in absence of the pump pulse.
-
Tables Table 1: Sample properties and maxima and linewidths of O-D absorption bands. zeolite
') d,
Si/Al
H/T
band
i&=
A;
b,
(cm-') 2620 42 2684 20
Ti ') (PSI 20 - 60 100-130
Y
2.0
.27
Mordenite
6.7
.13
LFd) 2649 HFd) 2664
40 22
30 - 95
ZSM-5
16
.06
2665
23
50
LF HF
Since we deuterated the reolites, D/T is f .7 x H/T. Full width at half maximum of the absorption bands. frequency-dependence; (estimated) lowest and highest observed values. deconvoluted with two gaussian lineshapes, in accordance with ref [lo].
Vibrational relaxation entails a redistribution of the energy in the O-D oscillator into accepting modes, for which the sum of energies is in resonance with the O-D vibration. In the experiments intense ( M 100 p J ) picosecond(20 ps) tunable infrared (2200 - 4500 cm-l) pulses are used; the laser has a bandwidth of approximately 11 cm-' at O-D absorption frequencies (a description of the experimental setup can be found in ref.[12]). A potential pitfall in our experiments is the steady-state heating of the sample due to the relatively large amounts of energy absorbed in the sample; relaxation times are intrinsically dependent on temperature. This heating would be wavenumber-dependent, as more energy is absorbed at the top of the absorption peak, than at the flanks. We checked for heating effects by changing the repetition rate of the pump-pulses. At rates of 5 HZ and lower no dependence was observed. All experiments were performed at a pump repetition rate of 5 Hz. Three different zeolites were investigated in our experiments: Y zeolite, Mordenite and ZSM-5. Our samples consist of pressed self-supporting discs of 3.5 to 6 mg/cma. Proton loaded zeolites were obtained by heating in zlacuo (for Y and ZSM-5 at least 1 h at 723 K, for Mordenite 1 h at 823 K) zeolites in which Na+ cations were exchanged by NH$ cations. In table 1 the properties of these zeolites and their infrared 0-D absorption bands can be found. Deuteration was achieved by adding 500 mbar of Da gas (Messer Griesheim, 99.7%) at 723 K and allowing exchange for 1 hour, resulting in approximately 70% exchange. Absorption spectra were recorded using a Perkin-Elmer 881 IR spectrometer.
495 Y-OD
2500
2600
Wovenumber (cm-')
2700
Fig.1 Absorption spectrum of the 0-D absorPtion bands of deuterated zeolite Y.The LF-Ped at 2620 on-' is caused bY tlvdmxyls Situated in the small sodalite Cag-9 the H F - P d bY hYdmXyls sticking into the supercage. Fits are doublegaussian fits to both the LF- and HF-peak.
MORDENITE-00
2620
2640
2660
Wovenumber
(crn-')
2680
Fig.2 Absorption spectrum of D-Mordenite. Dashed and dotted lines aregaussian deconve rs lutions corresponding to 0 - D ~ c i l ~ a t o situated in &ring cavities (LF,dotted) and 12-ring cavities (HF,dashed).
Results In Figures 1 and 2 the conventional infrared absorption spectra of zeolite D-Y and DMordenite are depicted. It is well known that in zeolite Y the LF peak is caused by hydroxyls located in the small cages, whereas the HF peak is caused by hydroxyls in the larger cavities. In a recent paper it was proposed that -similar to Y-zeolite- the absorption peak of the (deuterated) hydroxyls in Mordenite consists of two contributions; LF and HF subpeaks, resulting from hydroxyls in the small (8-ring) cages and in the larger 12-ring cages, respectively.[lO]The deconvolution was carried out for our absorption spectra and the results obtained (see Fig. 2) were similar to those found in ref. [lo]. Two typical results of pump-probe experiments for D-MOR are shown in figure 3. If the laser frequency is tuned to the left of the absorption band (at 2640 cm-'), a 2'1 of 36 ps is found. At the right of the absorption band (at 2672 cm-l) a much larger 2'1 is observed. This frequency dependence of 21' has been observed before for H-MOR.[4] In Figure 4 TI is plotted vs laser frequency through the absorption band of Mordenite. In zeolite D-Y, a similar behaviour of Tl for both the LF and HF band on scanning through the respective frequencies is observed (Fig 5). For deuterated ZSM-5the absorption spectrum and Ti vs. ij can be found in Fig. 6. The vibrational relaxation time T1seems to be independent of frequency for this zeolite.
496
Fig.3 Relative transmission of an infrared probe Fig.4 TI as a function of laser frequency for Morpulse In(T/To) (TOis the transmitted probe en- denite 0-D. TIis found to increase with freergy in absence of the pump pulse) as function quency. Laser bandwidth with gaussian shape of the delay between pump and probe pulses for has a full width at half height of I 1 em-'. zeolite Mordenite-0-D at two laser frequencies. The values for the energy decay times 21' obtained from single exponential fits (and the corresponding standard deviations) are denoted in the graph.The top of the absorption peak is situated at % 2662 m-'. The vibrational relaxation time is found to be dependent on frequency.
Calculations In general the expression for the time-dependent transmission for a homogeneously broadened absorption peak after a pump pulse at t = 0 reads:
T ( t )= Jm 0 dPL(P)exp[-A(P)(l - 2foexp -(t/Tl))]
(1)
where T ( t )is the transmitted energy of the probe pulse at delay t , L(P) the laser band, A(;) the absorption band, and fo the excited fraction of oscillators at t = 0. In eq. 1 the fraction of excited oscillators does not depend on frequency, as would be the case for inhomogeneous broadening; for inhomogeneously broadened bands the expression is somewhat more complicated. This simplification does not affect the results of our calculations, however. If the total absorption band can be split into two subbands, i.e. having a high-frequency (HF) and low-frequency (LF) component, A = ALF(P) A""(;), the transmission is given as the product of the two contributions: T ( t ) = T""(t) x T C F ( t ) . The two contributions can be calculated separately, with different TItimes. This procedure enables us to calculate T ( t ) , the transmission of the probe pulse, by solving eq. 1 numerically. We can perform this calculation at different wavenumbers and in this fashion we can calculate the experimentally observed 21' as a function of wavenumber. Input variables are AL"(P) and A*"(P) with corresponding T f F and Tf", optionally frequency dependent.
+
497 Y-OD
. 100
,100
. .
. .
I
.
. .
.
1
.
.
.
. ZSM-5 . . .-OD. I
- 0.1
-
C
, 3
-.
u
P
20 2600
2650
Wovenumber (cm-')
Fig.5 TI as a function of laser frequency for DY. Open circles are data points, dotted lines are simulations with T r F fixed at 128 pa. It is evident the simulations cannot adequately describe data.
2700 2640
2650
2660
2670
Wovenumber (cm-')
0 2680
Fig.6 Absorption spectrum and vibrational relaxation times for D-ZSM-6. Right m ' s . corresponds to absorption spectrum (dotted line), left axis to 7'1 times (black circles).
Discussion Measured TI times are essentially convolutions of the actual TI'S within the laser bandwidth. The effects of this convolution are minimized by investigating 0-D vibrational relaxation instead of 0-H, since the laser bandwidth is smaller by a factor of three at 0-D absorption frequencies compared to 0-H frequencies ( 11and 34 cm-l, respectively). This broadening is inherent in the generation of our infrared pulses.[l2] By contrast, on deuteration the absorption linewidths are only 30% narrower. A second reason to investigate the deuterated analogue of the acid zeolites is the fact that there is a substantial decrease in scattering of the pump pulse at 0-D absorption frequencies, enhancing the signal to noise ratio in our experiments. In a recent thorough investigation of 0-H relaxation in Y-zeolites with time resolved spectroscopy, it was proposed that an increase of TI with frequency suggests hydrogen bonding; a stronger H-bond weakens the original 0-H bond and consequently decreases the absorption frequency, while simultaneously enhancing the coupling to accepting modes, which results in a faster decay.[$] This would explain the decrease in TI at lower frequencies and the fact that the LF band is broadened compared to the HF band. In this study of 0-H-relaxation in Y-zeolites a frequency dependent TfFwas found, whereas TFFshowed no dispersion. A problem here, however, was the fact that the laser bandwidth exceeded the HF absorption linewidth. In order to check whether the frequency dependence of TI in Mordenite is due to two frequency independent contributions -an HF-contribution with TFFand an LF-contribution with TfF- and whether the observed frequency dependence of TIof the HF peak in Y-OD is caused by overlap with the LF peak and laset convolution, we simulated the experiments numerically.
498
For zeolite D-Y we calculated what the outcome of our experiments would be for TFF increasing with wavenumber as observed, and T,'IFindependent of wavenumber with a value of 123 ps. The results of this calulation are presented in Fig. 5 together with the data. The data suggest that the observed dependence of TI on frequency cannot be due to overlap of the LF and HF peaks and/or convolution of the spectrum with the laser band; The data cannot be fitted with frequency independent T:F. This is conclusive evidence for a frequency dependent TFF , contrary to what was previously found for 0-H[4].
MORDENITE-OD o experiment; vUSER = 2663 cm-'
0.6
n
t"
5
C
0.4
0.2
0 -100
0
100
200
300
Delay (PSI
Fig.T Results of a pumpprobe experiment at the top of the Mordenite 0 - D peak; Data (open circles) as in fig.1,single exponential fit (solid line) and simulation (dotted). The result of the simulations has a distinct non-exponential decay, contrary to the data.
'
The difference in Tl between HF and LF hydroxyls cannot be explained by a coupling to different accepting modes, since the difference in energy is too small for that. This difference must be explained through hydrogen-bonding, because the local environments are very similar. The frequency dependence of the TFFwould then point to a very weak hydrogen bond for these hydroxyls. The observed frequency-dependenceof TI in Mordenite could have two probable causes: It could be a consequence of the fact that the two types of hydroxyl causing the two absorption peaks are probed simultaneously in our experiment, each with its own, frequency independent TI. The obserued frequency dependence would then be caused by two frequency independent contributions. The second possibility is that there is a real frequency dependence in one or both of the sub-peaks. An effort to reproduce the dependence of TI
499
on wavenumber for Mordenite with T f F and TFFboth independent of frequency, resulted in unarbitrary values of 30 and 95 ps. A consequence of this deconvolution with frequency independent TYF and T:F is that intermediate calculated Tl times are a mixture of T f F and TFF,implying a double exponential decay. Experimentally we only observe single exponential decays, however; in Fig. 7 the experimental data and the result of the simulation are plotted. We can thus conclude that one or both of the two sub-peaks must exhibit frequency dependent behaviour with respect to TI, indicating hydrogen bonding. For D-ZSM-5 (spectrum and T1in Fig 6.) no significant dependence could be found. The protons in ZSM-5 are all situated in the 10-ring channel and it was shown recently that these protons are energetically very similar. [13] Apparently the differences are very small, and there is no hydrogen bonding, accounting for the absence of a frequency dependence.
Conclusion We observed frequency dependent T f D times in zeolite Y, both for the HF and LF peak, confirming the previously observed hydrogen bonding of the LF hydroxyls in the small sodalite cages. The frequency dependence of the HF hydroxyls is tentatively attributed to weak hydrogen bonding of these hydroxyls. For Mordenite this frequency dependence of TI was observed as well. It is asserted with the help of simulations that this dependence must be due to an intrinsic dependence of TI on C in one or both of the two hydroxyl species causing the absorption peak in Mordenite. The most likely candidate are the hydroxyls in the smaller 8-ring cages, causing the downshifted and broadened LF absorption peak, by analogy with Y zeolite. For ZSM-5, within experimental accuracy, no dependence of Tlwas observed, indicative of the absence of hydrogen bonding in this zeolite, in accordance with the pore size (10ring channel). This finding also supports the proposition that the hydroxyls in ZSM-5, although crystallographically different, are energetically very similar. Acknowledgements The work described in this paper is part of the research program of the Stichting Fundamenteel Onderzoek van de Materie (Foundation for Fundamental Research on Matter) and waa made possible by financial support from the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (Netherlands Organization for the Advancement of Research).
References [l]C. Hirose, Y. Goto, N. Akamatsu, J. Kondo and K. Domen, Surf. Sc. 283 (1993) 244. [2] J. Kubota, M. Furuki, Y. Goto, J. Kondo, A. Wada, K. Domen, and C. Hirose, Chern. Plays. Lett. 204 (1993) 273.
[3] M.P. Casassa, E.J. Heilweil, J.C. Stephenson, and R.R.Cavanagh, J. Electron Spectrosc. Relat. Phenom. 38 (1986) 257. [4] Marco J.P. Brugmans, Aart W. Kleyn, Ad Lagendijk, Wim P.J.H. Jacobs and Rutger
A. van Santen, Chern. Phys. Lett., 217 (1994) 117.
[5]B. N. J. Persson and R.Ryberg Phys. Rev. B. 40, 10273 (1985). [6]H. Stach, J. Jiinchen, H. Jerschkewitz, U. Lohse, B. Parlitz, B. Zibrowius and M. Hunger, J. Phys. Chem. 96, 8473 (1992). [7]H. Stach, J. Janchen, H. Jerschkewitz, U. Lohse, B. Parlitz, B. Zibrowius and M. Hunger, J. Phys. Chern. 96, 8480 (1992).
[a] M.Crocker, R.H.M. Herold, M.H.W.
Sonnemans, C.A. Emeis, A.E. Wilson, and J.N.
van der Moolen, J. Phys. Chem. 97, 432 (1993).
[9]M.H.W. Sonnemans, C. den Heijer and M. Crocker, J. Phys. Chern. 97, 440 (1993).
[lo] V.L. Zhoblenko, M.A. Makarova and J. Dwyer, J. Phys. Chern., 97 5962 (1993). [ll] I. Bank&s, J. Valyon, G.I.Kapustin, D. K d i , A.L. Klyachko and T.R. Brueva, Zeolites, 8 (1988) 189. [12]H.J. Bakker, P.C.M. Planken, L. Kuipers and A. Lagendijk, J. Chem. Phys. 94,1730 (1991). [13]G.J. Kramer and R.A. van Santen, J. Am. Chern. SOC. 115 (1993) 2882.
Zeolites and Related Microporous Maierials: Siaie of ihe Ari 1994 Studies in Surface Science and Catalysis, Vol. 84 0 1994 Elsevier Scicnce B.V. All rights reservcd.
493
Time Dependence of Vibrational Relaxation of Deuterated Hydroxyls in Acidic Zeolites. Mischa Bonn 1, Marco J.P. Brugmans', Aart W. Kleyn', Rutger A. van Santent and Ad Lagendijk* *I
* FOM-Institute for Atomic and Molecular Physics, Kruislaan 407, 1098 SJ Amsterdam,
Tbe Netherlands t Scbuit Institute of Catalysis, Eindboven University of Tecbnology, P.O. Box 513, 5600 MB Eindboven, The Netberlands Abstract. In this picosecond time-resolved study of zeolite hydroxyls we conclude from the 0-D vibrational population lifetime that the protons in Y zeolite as well as in Mordenite are hydrogen bonded. In deuterated ZSM-5no hydrogen-bonding was observed.
Introduction. Acidic zeolites are widely applied in hydrocarbon conversion reactions in the petrochemical industry. Their acidity is due to Brensted active hydroxyl groups, which are the catalytically active sites. The hydroxyl group is located between a silicon and an aluminum atom. In order to gain deeper insight into the dynamic properties of these sites, we have investigated the relaxation of vibrationally excited deuterated hydroxyl groups with time-resolved (picosecond) infrared saturation spectroscopy. Fkom our experiments we obtain 2'1, the vibrational lifetime, which cannot be obtained by conventional infrared spectroscopy, due to the fact that the width of the absorption bands is determined by inhomogeneous broadening. The population lifetimes are in the order of 100 ps [l-41, corresponding to a bandwidth of 0.05 cm-', whereas the absorption linewidths are in the order of 30 cm-'. The broadening is not due to dephasing; the behaviour of the width of the absorption peak with temperature can by no means be described by a theory of 2'2 broadening [5]). The relation between acidity, absorption frequency and local environment of the zeolite hydroxyl has been the subject of ongoing research. Lately, a great deal of attention has been focused on this question in zeolite Mordenite.[6-111 We investigated the wavenumber dependence of the 0-D vibrational population lifetime 21' for several zeolites, in order to obtain new information concerning the nature of the inhomogeneous broadening of the 0-D absorption peaks. Calculations were performed to support our findings. Experimental Section. In our pump-probe experiments a considerable fraction ( ~ 1 0 - 2 0 % )of the 0-D oscillators is excited from their v = 0 to 'u = 1vibrational state by a powerful short laser pulse. As a result the 0-D absorption is bleached on picosecond time scales; the transmission of light
494
of this wavelength through the sample increases temporarily. Hence the equilibration of the population-distribution can be monitored by measuring the transmission of a much weaker pulse -the probe pulse- whose time delay with respect to the pump-pulse can be varied. The decay of the pump-induced transmission with time is related to the vibrational lifetime of the excitation 21' , as: ln[T(t)/To] exp(-t/Tl), where T ( t )is the transmitted energy of the probe pulse at delay t and 0'2 is the transmitted probe energy in absence of the pump pulse.
-
Tables Table 1: Sample properties and maxima and linewidths of O-D absorption bands. zeolite
') d,
Si/Al
H/T
band
i&=
A;
b,
(cm-') 2620 42 2684 20
Ti ') (PSI 20 - 60 100-130
Y
2.0
.27
Mordenite
6.7
.13
LFd) 2649 HFd) 2664
40 22
30 - 95
ZSM-5
16
.06
2665
23
50
LF HF
Since we deuterated the reolites, D/T is f .7 x H/T. Full width at half maximum of the absorption bands. frequency-dependence; (estimated) lowest and highest observed values. deconvoluted with two gaussian lineshapes, in accordance with ref [lo].
Vibrational relaxation entails a redistribution of the energy in the O-D oscillator into accepting modes, for which the sum of energies is in resonance with the O-D vibration. In the experiments intense ( M 100 p J ) picosecond(20 ps) tunable infrared (2200 - 4500 cm-l) pulses are used; the laser has a bandwidth of approximately 11 cm-' at O-D absorption frequencies (a description of the experimental setup can be found in ref.[12]). A potential pitfall in our experiments is the steady-state heating of the sample due to the relatively large amounts of energy absorbed in the sample; relaxation times are intrinsically dependent on temperature. This heating would be wavenumber-dependent, as more energy is absorbed at the top of the absorption peak, than at the flanks. We checked for heating effects by changing the repetition rate of the pump-pulses. At rates of 5 HZ and lower no dependence was observed. All experiments were performed at a pump repetition rate of 5 Hz. Three different zeolites were investigated in our experiments: Y zeolite, Mordenite and ZSM-5. Our samples consist of pressed self-supporting discs of 3.5 to 6 mg/cma. Proton loaded zeolites were obtained by heating in zlacuo (for Y and ZSM-5 at least 1 h at 723 K, for Mordenite 1 h at 823 K) zeolites in which Na+ cations were exchanged by NH$ cations. In table 1 the properties of these zeolites and their infrared 0-D absorption bands can be found. Deuteration was achieved by adding 500 mbar of Da gas (Messer Griesheim, 99.7%) at 723 K and allowing exchange for 1 hour, resulting in approximately 70% exchange. Absorption spectra were recorded using a Perkin-Elmer 881 IR spectrometer.
495 Y-OD
2500
2600
Wovenumber (cm-')
2700
Fig.1 Absorption spectrum of the 0-D absorPtion bands of deuterated zeolite Y.The LF-Ped at 2620 on-' is caused bY tlvdmxyls Situated in the small sodalite Cag-9 the H F - P d bY hYdmXyls sticking into the supercage. Fits are doublegaussian fits to both the LF- and HF-peak.
MORDENITE-00
2620
2640
2660
Wovenumber
(crn-')
2680
Fig.2 Absorption spectrum of D-Mordenite. Dashed and dotted lines aregaussian deconve rs lutions corresponding to 0 - D ~ c i l ~ a t o situated in &ring cavities (LF,dotted) and 12-ring cavities (HF,dashed).
Results In Figures 1 and 2 the conventional infrared absorption spectra of zeolite D-Y and DMordenite are depicted. It is well known that in zeolite Y the LF peak is caused by hydroxyls located in the small cages, whereas the HF peak is caused by hydroxyls in the larger cavities. In a recent paper it was proposed that -similar to Y-zeolite- the absorption peak of the (deuterated) hydroxyls in Mordenite consists of two contributions; LF and HF subpeaks, resulting from hydroxyls in the small (8-ring) cages and in the larger 12-ring cages, respectively.[lO]The deconvolution was carried out for our absorption spectra and the results obtained (see Fig. 2) were similar to those found in ref. [lo]. Two typical results of pump-probe experiments for D-MOR are shown in figure 3. If the laser frequency is tuned to the left of the absorption band (at 2640 cm-'), a 2'1 of 36 ps is found. At the right of the absorption band (at 2672 cm-l) a much larger 2'1 is observed. This frequency dependence of 21' has been observed before for H-MOR.[4] In Figure 4 TI is plotted vs laser frequency through the absorption band of Mordenite. In zeolite D-Y, a similar behaviour of Tl for both the LF and HF band on scanning through the respective frequencies is observed (Fig 5). For deuterated ZSM-5the absorption spectrum and Ti vs. ij can be found in Fig. 6. The vibrational relaxation time T1seems to be independent of frequency for this zeolite.
496
Fig.3 Relative transmission of an infrared probe Fig.4 TI as a function of laser frequency for Morpulse In(T/To) (TOis the transmitted probe en- denite 0-D. TIis found to increase with freergy in absence of the pump pulse) as function quency. Laser bandwidth with gaussian shape of the delay between pump and probe pulses for has a full width at half height of I 1 em-'. zeolite Mordenite-0-D at two laser frequencies. The values for the energy decay times 21' obtained from single exponential fits (and the corresponding standard deviations) are denoted in the graph.The top of the absorption peak is situated at % 2662 m-'. The vibrational relaxation time is found to be dependent on frequency.
Calculations In general the expression for the time-dependent transmission for a homogeneously broadened absorption peak after a pump pulse at t = 0 reads:
T ( t )= Jm 0 dPL(P)exp[-A(P)(l - 2foexp -(t/Tl))]
(1)
where T ( t )is the transmitted energy of the probe pulse at delay t , L(P) the laser band, A(;) the absorption band, and fo the excited fraction of oscillators at t = 0. In eq. 1 the fraction of excited oscillators does not depend on frequency, as would be the case for inhomogeneous broadening; for inhomogeneously broadened bands the expression is somewhat more complicated. This simplification does not affect the results of our calculations, however. If the total absorption band can be split into two subbands, i.e. having a high-frequency (HF) and low-frequency (LF) component, A = ALF(P) A""(;), the transmission is given as the product of the two contributions: T ( t ) = T""(t) x T C F ( t ) . The two contributions can be calculated separately, with different TItimes. This procedure enables us to calculate T ( t ) , the transmission of the probe pulse, by solving eq. 1 numerically. We can perform this calculation at different wavenumbers and in this fashion we can calculate the experimentally observed 21' as a function of wavenumber. Input variables are AL"(P) and A*"(P) with corresponding T f F and Tf", optionally frequency dependent.
+
497 Y-OD
. 100
,100
. .
. .
I
.
. .
.
1
.
.
.
. ZSM-5 . . .-OD. I
- 0.1
-
C
, 3
-.
u
P
20 2600
2650
Wovenumber (cm-')
Fig.5 TI as a function of laser frequency for DY. Open circles are data points, dotted lines are simulations with T r F fixed at 128 pa. It is evident the simulations cannot adequately describe data.
2700 2640
2650
2660
2670
Wovenumber (cm-')
0 2680
Fig.6 Absorption spectrum and vibrational relaxation times for D-ZSM-6. Right m ' s . corresponds to absorption spectrum (dotted line), left axis to 7'1 times (black circles).
Discussion Measured TI times are essentially convolutions of the actual TI'S within the laser bandwidth. The effects of this convolution are minimized by investigating 0-D vibrational relaxation instead of 0-H, since the laser bandwidth is smaller by a factor of three at 0-D absorption frequencies compared to 0-H frequencies ( 11and 34 cm-l, respectively). This broadening is inherent in the generation of our infrared pulses.[l2] By contrast, on deuteration the absorption linewidths are only 30% narrower. A second reason to investigate the deuterated analogue of the acid zeolites is the fact that there is a substantial decrease in scattering of the pump pulse at 0-D absorption frequencies, enhancing the signal to noise ratio in our experiments. In a recent thorough investigation of 0-H relaxation in Y-zeolites with time resolved spectroscopy, it was proposed that an increase of TI with frequency suggests hydrogen bonding; a stronger H-bond weakens the original 0-H bond and consequently decreases the absorption frequency, while simultaneously enhancing the coupling to accepting modes, which results in a faster decay.[$] This would explain the decrease in TI at lower frequencies and the fact that the LF band is broadened compared to the HF band. In this study of 0-H-relaxation in Y-zeolites a frequency dependent TfFwas found, whereas TFFshowed no dispersion. A problem here, however, was the fact that the laser bandwidth exceeded the HF absorption linewidth. In order to check whether the frequency dependence of TI in Mordenite is due to two frequency independent contributions -an HF-contribution with TFFand an LF-contribution with TfF- and whether the observed frequency dependence of TIof the HF peak in Y-OD is caused by overlap with the LF peak and laset convolution, we simulated the experiments numerically.
498
For zeolite D-Y we calculated what the outcome of our experiments would be for TFF increasing with wavenumber as observed, and T,'IFindependent of wavenumber with a value of 123 ps. The results of this calulation are presented in Fig. 5 together with the data. The data suggest that the observed dependence of TI on frequency cannot be due to overlap of the LF and HF peaks and/or convolution of the spectrum with the laser band; The data cannot be fitted with frequency independent T:F. This is conclusive evidence for a frequency dependent TFF , contrary to what was previously found for 0-H[4].
MORDENITE-OD o experiment; vUSER = 2663 cm-'
0.6
n
t"
5
C
0.4
0.2
0 -100
0
100
200
300
Delay (PSI
Fig.T Results of a pumpprobe experiment at the top of the Mordenite 0 - D peak; Data (open circles) as in fig.1,single exponential fit (solid line) and simulation (dotted). The result of the simulations has a distinct non-exponential decay, contrary to the data.
'
The difference in Tl between HF and LF hydroxyls cannot be explained by a coupling to different accepting modes, since the difference in energy is too small for that. This difference must be explained through hydrogen-bonding, because the local environments are very similar. The frequency dependence of the TFFwould then point to a very weak hydrogen bond for these hydroxyls. The observed frequency-dependenceof TI in Mordenite could have two probable causes: It could be a consequence of the fact that the two types of hydroxyl causing the two absorption peaks are probed simultaneously in our experiment, each with its own, frequency independent TI. The obserued frequency dependence would then be caused by two frequency independent contributions. The second possibility is that there is a real frequency dependence in one or both of the sub-peaks. An effort to reproduce the dependence of TI
499
on wavenumber for Mordenite with T f F and TFFboth independent of frequency, resulted in unarbitrary values of 30 and 95 ps. A consequence of this deconvolution with frequency independent TYF and T:F is that intermediate calculated Tl times are a mixture of T f F and TFF,implying a double exponential decay. Experimentally we only observe single exponential decays, however; in Fig. 7 the experimental data and the result of the simulation are plotted. We can thus conclude that one or both of the two sub-peaks must exhibit frequency dependent behaviour with respect to TI, indicating hydrogen bonding. For D-ZSM-5 (spectrum and T1in Fig 6.) no significant dependence could be found. The protons in ZSM-5 are all situated in the 10-ring channel and it was shown recently that these protons are energetically very similar. [13] Apparently the differences are very small, and there is no hydrogen bonding, accounting for the absence of a frequency dependence.
Conclusion We observed frequency dependent T f D times in zeolite Y, both for the HF and LF peak, confirming the previously observed hydrogen bonding of the LF hydroxyls in the small sodalite cages. The frequency dependence of the HF hydroxyls is tentatively attributed to weak hydrogen bonding of these hydroxyls. For Mordenite this frequency dependence of TI was observed as well. It is asserted with the help of simulations that this dependence must be due to an intrinsic dependence of TI on C in one or both of the two hydroxyl species causing the absorption peak in Mordenite. The most likely candidate are the hydroxyls in the smaller 8-ring cages, causing the downshifted and broadened LF absorption peak, by analogy with Y zeolite. For ZSM-5, within experimental accuracy, no dependence of Tlwas observed, indicative of the absence of hydrogen bonding in this zeolite, in accordance with the pore size (10ring channel). This finding also supports the proposition that the hydroxyls in ZSM-5, although crystallographically different, are energetically very similar. Acknowledgements The work described in this paper is part of the research program of the Stichting Fundamenteel Onderzoek van de Materie (Foundation for Fundamental Research on Matter) and waa made possible by financial support from the Nederlandse Organisatie voor Wetenschappelijk Onderzoek (Netherlands Organization for the Advancement of Research).
References [l]C. Hirose, Y. Goto, N. Akamatsu, J. Kondo and K. Domen, Surf. Sc. 283 (1993) 244. [2] J. Kubota, M. Furuki, Y. Goto, J. Kondo, A. Wada, K. Domen, and C. Hirose, Chern. Plays. Lett. 204 (1993) 273.
[3] M.P. Casassa, E.J. Heilweil, J.C. Stephenson, and R.R.Cavanagh, J. Electron Spectrosc. Relat. Phenom. 38 (1986) 257. [4] Marco J.P. Brugmans, Aart W. Kleyn, Ad Lagendijk, Wim P.J.H. Jacobs and Rutger
A. van Santen, Chern. Phys. Lett., 217 (1994) 117.
[5]B. N. J. Persson and R.Ryberg Phys. Rev. B. 40, 10273 (1985). [6]H. Stach, J. Jiinchen, H. Jerschkewitz, U. Lohse, B. Parlitz, B. Zibrowius and M. Hunger, J. Phys. Chem. 96, 8473 (1992). [7]H. Stach, J. Janchen, H. Jerschkewitz, U. Lohse, B. Parlitz, B. Zibrowius and M. Hunger, J. Phys. Chern. 96, 8480 (1992).
[a] M.Crocker, R.H.M. Herold, M.H.W.
Sonnemans, C.A. Emeis, A.E. Wilson, and J.N.
van der Moolen, J. Phys. Chem. 97, 432 (1993).
[9]M.H.W. Sonnemans, C. den Heijer and M. Crocker, J. Phys. Chern. 97, 440 (1993).
[lo] V.L. Zhoblenko, M.A. Makarova and J. Dwyer, J. Phys. Chern., 97 5962 (1993). [ll] I. Bank&s, J. Valyon, G.I.Kapustin, D. K d i , A.L. Klyachko and T.R. Brueva, Zeolites, 8 (1988) 189. [12]H.J. Bakker, P.C.M. Planken, L. Kuipers and A. Lagendijk, J. Chem. Phys. 94,1730 (1991). [13]G.J. Kramer and R.A. van Santen, J. Am. Chern. SOC. 115 (1993) 2882.