J. Phys. Chem. 1993,97, 339-343
339
Growth of Silver Particles in Aqueous Solution: Long-Lived “Magic” Clusters and Ionic Strength Effects B. G. Ershov,’ E. Janata, and A. Henglein’ Hahn-Meitner-Institut Berlin GmbH, Bereich S,lo00 Berlin 39, FRG Received: September 16, I992
Pulse radiolytic reduction of silver ions leads to an oligomeric cluster absorbing at 295 nm and most intensely and sharply at 325 nm. The cluster has a half-life of >5 min in scrupulously clean reaction vessels. During its decay, larger metallic particles absorbing around 380 nm are produced. The cluster is long-lived only in the presence of excess Ag+ ions. A second cluster absorbing at 345 nm is also formed, which lives for about 100 s. The 325-nm cluster is oxidized by 02,but larger particles are also formed during the oxidation. Tetranitromethane reacts with the cluster to form the nitroform anion, from which the absorption coefficient per Ag atom is calculated to be 2 X 104 M-1 cm-l. The formation of the cluster is accelerated by NaC104; the Bjerrum-Branstedt evaluation of this effect shows that a precursor of the cluster carries three elementary charges. The formation of larger metallic particles is even more strongly accelerated by NaC104. In the presence of M NazS04, the formation and decay of the cluster occur practically at the diffusion-controlled rate. From the rate of the buildup under these conditions, it is estimated that the cluster contains eight reduced silver atoms. The effects of ionic strength on the growth of the clusters and metallic particles are compared to the salt effects in slow andfast coagulation of colloids.
The reduction of silver ions in aqueous solution generally yields colloidal silver with particle diameters of several nanometers. These particles have metallic properties, one of these properties being the presence of an intense plasmon absorption band in the 380-400-nm range which is caused by a collective oscillation of the electron gas. When the reduction is carried out in a pulse experiment (by applying a short high-energy electron pulse to the silver salt solution), intermediate oligomeric clusters can be dettcted.1,2 Their absorption bands lie at shorter wavelengths than the plasmon band of the larger metallic particles. The buildupof the cluster bands occurs within some tenths of a second after the pulse, and their lifetime amounts to a few seconds. In the present paper, new pulse radiolysis experiments are reported in which pulse trains are used for the preparation of cluster solutions. A photodiode array is used to detect intermediates in the millisecond to minute time regime. The lifetime of the oligomeric clusters is found to be very much longer than in the earlier experiments. Under certain conditions the clusters formed by pulse irradiation were so long-lived that they could subsequently be detected in a conventional spectrophotometer. The effects of ionic strength on the rate of buildup and disappearance of the clusters and of the appearance of larger metallic particles are also reported. It has to be emphasized that the experimentsoccurred under conditionswhere there was always an excess of Ag+ ions in the solution. In a succeeding paper, the effects will be described which occur when all the Ag+ ions are reduced by the pulse; i.e., no Ag+ ions are present that could stabilize the silver clusters.
Experimental Section The setup for conventional pulse radiolysis has already been described (3.8-MeV electrons, 1.5-ps pulse width). In some of the exFriments, pulse trains of high-energy electrons were used to prepare colloids;this method has also been described recentl~.~ The intensity of each pulse was monitored as reported previ~usly.~ ~~
* To whom correspondence should be addressed.
‘On leave of absence from the Institute of Physical Chemistry of the Academy of Sciences, Moscow,Russia.
0022-3654f 58f 2091-0339$04.00f 0
In order to rapidly record absorption spectra in the millisecond to minute range after a pulse, the monochromator/photodetector unit was replaced by a multichannel detector (Polytec), which consists of a photodiode array and an optical grating. The analyzing light is spread by the grating onto the array (1024 elements). A spectrum can in principle be recorded from 220 to 800 nm with a resolution of better than 1 nm, the measuring time being about 25 ps per element used. However, the intensity of the analyzing light below 260 nm was too low to allow recording at shorter wavelengths. The multichannel detector is connectedto a HP 382 workstation via a IEEE-488 bus. The optical absorbance base line is recorded before the experiment. Irradiation occurs with one single pulse or a train of pulses. The solutions were deaerated either by evacuation (see the vessel in Figure 1 of ref 3) or by bubbling with argon.
Results
Cluster Formation a d Stability. In the experiments of Figure 1,a 1 X 10-4M AgC104s01utionwhich contained 0.1 M 2-propanol was irradiated with pulse trains of different length. In each pulse, 1 X 10-5 M reducing radicals (hydrated electrons and organic radicals) were produced. After irradiation, the spectrum of the solution was recorded in a conventional spectrophotometer. The time required to transport the solution from the irradiation room to the spectrophotometer and to record the spectrum was about 2 min. Despite this long time, the solutioncontained strongcluster absorption bands at 295 and 325 nm. When between 1 and 10 pulses were applied, the bands increased in intensity. After the application of 20 pulses, these bands were no longer present in the spectrum, but the plasmon band of larger metallic particles was present. The plasmon band was first located at 370 nm (20 pulses) and then at 380 nm (100 pulses). There is a rather sharp transition from the oligomeric clusters absorbing at 295 and 325 nm to the larger metallic particles in the range between 10 and 20 pulses. After 10 pulses almost all the Ag+ ions are reduced. It thus seems that the intermediate cluster is long-lived only in the presence of a certain amount of excess Ag+ ions in the solution. The cluster bands disappear within minutes as can be seen from Figure 2. The rates of disappearance of the absorption (6
1993 American Chemical Society
340 The Journal of Physical Chemistry, Vol. 97, No. 2, 1993 1.5
J
Ershov et al.
1
1.0
c
n
e 0 II)
n
"0.5 I
I
I
200
1
300
I
I
400
500
A Inml Figure 1. Spectrum of a solution which was irradiated with pulse trains. The number of pulses applied is given on the curves. For each pulse I X 10-5 M reducing radicals (hydrated electrons and organic radicals) were generated. The interval between the pulses was 3.3 ms. The evacuated solutioncontained 1 X lo-' M AgClOd and 0.1 M 2-propanol. Irradiation vessel as in Figure 1 of ref 3.
300
I
I
300
I
I
400
I
I 500
A Inml Figure 2. Spectrum of a cluster solution at various times of aging. The cluster solution was prepared by irradiating the solution of Figure 1 with five pulses. The spectrum was recorded in a conventional spectrophotometer. To obtain the lifetime of the cluster, 2 min (the time for transporting the solution from the irradiation room to the spectrophotometer and recording the spectrum) has to be added to the times given in the figure.
bands at 295 and 325 nm are qual, which may be taken as an indication that both bands belong to the same cluster. During the disappearanceof the cluster bands, the plasmon band of larger metallic particles is built up, which indicates that the 325-nm cluster is the precursor to the larger particles. The plasmon band present after 7 min has its maximum at 400 nm, i.e., at a longer wavelength than in Figure 1. This is due to the formation of agglomerated particles as no polymeric stabilizer was present in the solution. In fact, when the solution was aged for longer times than shown in Figure 2, the intensity of the plasmon band decreased and the band broadened due to continued agglomeration. The long-lived cluster could only be observed when the irradiation vessel was carefully cleaned with KMn04 plus concentrated HC104 toremoveall traccsof impurities. Otherwise, the product of irradiation was colloidal silver. It was found that even micromolar concentrations of metal ions, such as Cd2+ and Tl+, as well as organic electron acceptors, such as nitrobenzene, catalyze the transformation of the cluster into the larger colloidal particles. In the experiment of Figure 2, the lifetime of the cluster was about 5 min. Occasionally, lifetimes up to 15 min were observed, the reproducibility being low, which is attributed to traces of impurities. As has been shown previously by conventionalpulse radiolysis, the cluster is built up in the 0.1-s range, if the solution is exposed
380
A Inml Figure 3. Buildup of the cluster absorption. Spectrum at different times after the pulse as rccorded with the photodiode array. The solution was irradiated with a single pulse producing 2 X IO-' M reducing radicals. Solution as in Figure 1.
"300
0' 200
340
400
X Inml Figure 4. Absorption spectrum at various times after the irradiation with four pulses. Spectrum recorded with the photodiode array. Solution as in Figure 1.
to one strong pulse of high-energy electrons.' In the experiment of Figure 3, the photodide array was used to follow the buildup of the absorption of the clusters. One pulse producing 2 X 10-5 M reducing radicals was applied, and the spectrum was recorded after various lengths of time. The first spectrum, taken at 0.01 s, contains the 325-nm cluster band with low intensity and a broad absorption band around 280 nm (which masks the 295-nm cluster band) as well as a weak band at 355 nm. As the 295- and 325-nm bands increase with time, the 280- and 355-nm bands decrease;the two latter bands have to beattributed to twoclusters (possibly Ag42+,280 nm,IJ and 2 4 + , 355 nmz) which are precursors of the 325-nm cluster. After 4.8 s, the absorption of the 325-nm cluster reached a maximum. The half-life of 0.4 s for the buildup of the cluster agrees with the value found earlier.lJ The cluster is more rapidly built up at higher doses, Le., higher concentrations of reduced silver. In the experiment of Figure 4, the solution was exposed to four pulses. At detection times shorter than those in the experiment in Figure 1, a broad absorption band which peaks at 345 nm is recorded as well as the 325-nm band. The 345-nm band decays within about 100 s whereas the 325-nm band is still present. One concludes that, under the conditions in Figure 4, a new cluster with a shorter lifetime is present; this cluster could not bedetected in the experiments where the spectrum was measured in the conventional spectrophotometer (Figure l), because the cluster decays during the transfer of the irradiated solution to the spectrophotometer. Oxidation of the 3 2 J m Cluster. The 325-nm cluster is oxidized by oxygen. In the experiment of Figure 5, this oxidation was carried out very slowly: the vessel containing the cluster solution was exposed to air without shaking the solution. The
The Journal of Physical Chemistry, Vol. 97, No. 2, 1993 341
Growth of Ag Particles in Aqueous Solution
0.10
-0.08 -5 v
u
m
0.06
f 0 VI
% 0.04
u 300 400
‘350
350 450 A lnml Figwe 5. Spectrum of a cluster solutionbefore and after two time periods of contact with air.
fv) I
-I+ 10 I
0.1
I
I
I
I
I
I _
0.2
INaCI0,l’’’ Figure 6. Reciprocal half-life of cluster formation as a function of concentration of added NaC104 in a Brbnstedt-Bjerrum plot.
oxygen could therefore penetrate only via diffusionand convection as the solution stood in the spectrophotometer. The spectrum at various times as recorded in a conventional spectrophotometer is shown in Figure 5. One can see that, as the cluster bands fade away, a weak band at 400 nm develops which finally decreases and strongly broadens. This latter band is ascribed to larger metallic particles. Under the conditions applied, silver atoms are not only oxidized but also partly agglomerate to yield larger particles (which are less sensitive toward 02). M tetranitromethane to a cluster Upon the addition of solution, thecluster absorptionsdisappeared instantaneously,and the 350-nm band of the anion of nitroform was present: Ag,
+ (m/Z)C(NO,),
-
mAg+ + (m/2)C(NO2);
+
(m/2)NO,- (1) From the absorption coefficient of 1.4 X 104 M-l cm-I of this anion, an absorption coefficient of 2 X 104 M-I cm-1 per silver atom of the cluster was calculated. This value is close to what has been estimated previously.6 It constitutes a lower limit as it cannot be excluded that the cluster solution contained a small concentration of larger particles which also react with tetranitromethane. I d c Stnaqtb Effects. When sodium perchlorate was present in the solution, the buildup of the cluster occurred more rapidly.2 In fact, it became necessary to apply conventional detection techniques to record the absorption spectrum at shorter times. In Figure 6, the reciprocal first half-life for the buildup of the 325nm band is plotted on a logarithmic scale versus the square root of the salt concentration. From the slope of the straight line obtained and using the Branstedt-Bjerrum relation, log 1/ r = log 1/ T O 1.02zazJ, one calculates for the product ZaZb of the
+
400
500
F i p e 7 . Absorptionspectrumat varioustimesafter thepulseofasolution which contained 0.05 M NaC104. Solution as in Figure 1.
5. I!-2
100
0
300
A lnml
INaCIOJ 5.,IO‘3
3
0.02
reacting species a value of 9. Thus, it appears as if the ratedetermining step of the cluster buildup is a reaction between precursors carrying three elementary charges. In Figure 7 the absorption spectrum of a pulsed solution, which contained 0.05 M NaCIO4, is shown at various times after the pulse. Before the clusters are formed (spectrum at 0.5 ms), the absorption bands of precursors are seen at 270 and 355 nm. The cluster absorption bands at 295 and 325 nm are already fully developed after 40 ms. At this time, the solution also contains some larger particles as can be seen from the broad absorption maximum around 390 nm. At 1 s after the pulse, the spectrum contains the absorption band of larger metallic particles (close to 400 nm) and only a very small residual absorption band of the cluster. Obviously, the presence of NaC104 accelerates not only the formation of the cluster but also its conversion into larger particles. The quantitative investigation of the rate of colloid formation as a function of the NaC104 concentration revealed that the acceleration is even stronger than that of the cluster formation(Figure 6). One may compare the effectiveacceleration of the formation of the larger particles with the salt effect on the coagulation of colloids. It is known from many experiments in conventional colloid science that the rate of coagulation increases with a high power of the salt concentration. As reported previously,’ sulfate ions strongly accelerate both the formationof thecluster anditsconversion intolarger particles. Figure 8 shows the absorption spectrum of a solution to which 1 X 10-0 M sodium sulfate had been added. It can be seen that the colloid is already fully developed after 10 ms. The cluster band at 325 nm is present only weakly at 0.8 ms; the half-life time for the buildup of the 325-nm absorption was 300 ps. The spectrum at 0.8 ms already contains the absorption of some larger particles (second band at 375 nm). The precursor absorption bands at 280 and 355 nm are seen after 0.1 ms in Figure 8.
Discussion Most surprising is the long cluster lifetime of >5 min of the cluster which is formed in the electron pulse irradiation of a silver ion solution (Figure 2). The absorption bands of the cluster are rather sharp, which indicates that just one cluster is prtsent and not a mixture of clusters of various sizes. When the cluster disappears larger aggregates which have metallic character are produced (Figure 2). The 325-nm cluster must have great stability, possibly due to a “magic” agglomeration number. In addition, the cluster must be stabilized by an ionic charge and double layer as it disappears more rapidly in solutions of higher ionic strength. The cluster lives for days if it is formed in the presence of a polyanion.68
342
The Journal of Physical Chemistry, Vol. 97, No. 2, 1993 0.10
z
1 I
.':.r (Oms
. .
\
Ershov et al. one may expect that Ag+ ions are split off during the course of the agglomeration processes. It also appears plausible to regard only those reactions, where the products would have an even number of electrons, such as
I I
I
300
400
500
A lnml Figures. Absorptionspcctrumatvarioustimesafter thepulseofa solution M sodium sulfate. Solution as in Figure 1. which contained 1 X
The absorption coefficient per Ag atom in the cluster (2 X lo4 M-l cm-I) is comparable to that of larger metallic particles which possess the plasmon band at 380 nm. When the sharp absorption band of the cluster was observed for the first time, it was thought that its position at rather short wavelengths was an indication that the nature of the optical transition occurring was quite different from the plasmon oscillation in the larger metallic particles. In fact, it was proposed that a charge-transfer-tosolvent transition was responsible for the intense absorption band and that the cluster had an odd number of electrons.6 However, as stabilized cluster solutions do not show ESR activity, the electron number must be even. Giant absorption bands of rather small clusters in molecular beams in vacuo have been reported during the past few years in the case of alkali metals.9 It appears plausible that the 295- and 325-nm bands of our silver cluster constituteanother examplefor collective electron excitation taking place even at cluster sizes where the typical metal properties are not yet fully developed. Most recently, rather sharp absorption bands have been reported for mass-separated neutralsilver clusters in argon, which lie close to the 325-nm wavelength of our cluster."J Pulse radiolysis experiments on the reduction of lead ions have also revealed that rather small Pb clusters have a giant absorption band. I I The exact size of the cluster is not yet known. We try now to estimate it from kinetic data on the rate of its formation. First, it should be recalled that the rate at which the cluster builds up is very low under low ionic strength conditions (Figure 2), the half-life being 0.4 s. The strong dependence of the rate of buildup on the ionic strength (as adjusted by NaC104) indicates that a precursor of the cluster must have an elementary charge close to 3 (Figure 6). The first steps of the reduction of Ag+ by hydrated electrons are well-known: Ag+
-
+ eaq-
Ago + Ag+
Ago Ag, +
k = 3 X 10" M-' s-' k =5
X
lo9 M-'s-I
(2)
(3)
These reactions are complete in less than 3 ps under our experimental conditions. The following steps of agglomeration require much more time, the first one being Agq2+is a species that already carries a relatively high ionic charge. Whereas reaction 4 is known from pulse radiolysis observations, the following steps of agglomeration can only be described in a speculative manner. As it appears improbable that species with a large number of positive charges are stable,
Ag;
+ 5Ag'
(7)
If one assumes that the various steps of agglomeration occur with the same specific rate, the buildup time of a cluster formed in the nth agglomeration step (containing2" reduced silver atoms) would be roughly E:,,n7, where 7 = l/cRk is the half-time of the first agglomeration step (eq 4: CR = concentration of r e d u d silver atoms, k = specific rate of agglomeration). If one regards conditions where the agglomerations occur at a diffusioncontrolled rate, k is expected to amount to close to lo9 M-l s-'. (A slight modification to the above mechanism in which the organic radical generated in the presence of 2-propanol reduces part of Ag2+ to form Agz (which then participates in the agglomerations) does not change the principles of our consideration.) The enormous acceleration of agglomeration processes by 10-4 M sulfate (Figure 7) cannot be understood in terms of an ionic strength effect. One has to postulate complex formation between the charged precursors of the 325-nm cluster and sulfate anions. Under these conditions, the agglomerations may be expected to occur at a diffusion-controlled rate, as there do not exist strong repulsive forces between the particles. The buildup of the 325nm cluster was 300 ps (Figure 8). With k 3 2 X 109 M-1 s-I and CR = 1 X 10-5 M,one calculates a buildup time close to this value with a n value of 3. This would mean that the 325-nm cluster consists of eight reduced silver atoms. It should be remembered that our kinetic approach for estimatingtheagglomerationnumber is very simple; however, it seems safe to say that the 325-nm cluster consists of close to and probably less than 10 atoms. The cluster is possibly complexed by an Ag+ to yield a charged species, which would explain why the cluster is so stable even in the absence of a stabilizing polymer and becomes unstable when there are no excess Ag+ ions in the solution. The formation of larger particles from the cluster is even more stronglyaccelerated by sulfate than its formation. In the presence of sulfate, the process offast coagulation which is well understood for large colloidal particles in conventional colloid science may be invoked to explain the fast acceleration.I2 One is probably dealing with a continuous transition from complex formation in the case of molecular species (such as Ag2+)to specific adsorption of SO4*-in the case of the larger particles, the accelerating effect becoming more and more pronounced with increasing size of the agglomerating species. A precursor to the 325-nm cluster absorbs at 355 nm (Figure 3,O.Ol s; Figure 7,O.S ms; Figure 8,O. 1 ms). Besides the 325-nm cluster to which the structure of Ags is attributed, a cluster absorbing at 345 nm is formed. This latter cluster is short-lived (100 s), but it can be stabilized for hours when polyphosphate is presente6 The stabilized 345-nm cluster is also shorter-lived than the stabilized 325-nm cluster.6 It is also more reactive toward oxidizing reagents than the 325-nm ~1uster.I~ The nature of the 345-nm cluster is not yet clear; it could be the Ags cluster in another charge state (carrying a negative charge?) or a cluster of another "magic" number. Final Remarks
The agglomeration processes that occur during the reduction of a metal in solution are of special interest in the case of silver.
Growth of Ag Particles in Aqueous Solution It is the only case where long-lived intermediate clusters can be observed even in the absence of a polymeric stabilizer that have not yet fully acquired metallic properties. Twelve years ago, the first evidence was obtained for rather small metal agglomerates ( n < 15) possessing absorption bands like larger metallic particles.14 In the present paper, this effect is reinvestigated to shed light on the atom metal transition in a more detailed manner. The changes in the chemical properties with changing cluster size have previously been reported.6v7J3 In the present paper, the growth of the aggregates under various ionic strength conditions is compared with the effects of added salts on colloids. The effect of increasing ionic strength by a 1:l electrolyte (NaC104)on the transition from nonmetallic to metallic particles is described by the kinetic salt effect (Bjerrum-Brhstedt) at the early stages of coalescence, but for the larger particles by slow coagularionmodels as used in colloid chemistry. A 1:2 electrolyte, Na2S04,creates conditions of particle growth comparable tofast coagulation in colloid science.
-
Acknowledgment. The authors thank Dr. Paul Mulvaney for helpful discussions and reading the manuscript. References and Notes (1) Mulvaney, P.; Henglein, A. Chem. Phys. Lerr. 1990, 168, 391.
The Journal of Physical Chemistry, Vol. 97, No. 2, 1993 343 (2) (a) Enhov, B. G.; Sukhov, N. L.; Troitskii, D. A. Zv. Akad. Nauk. SSSR Ser. Khim. 1989, No. 8,1930; High Energy Chem. 1991,25,176. (b) Ershov, B. G.; Sukhov, N. L.; Troitskii, D. A. Radiar. Phys. Chem. 1992,39, 127. (3) Henglein, A.; GutiCrrez,M.; Janata, E.; Ershov, E.G. J. Phys. Chem. 1992, 96, 4590. (4) Janata, E. Radiar. Phys. Chem. 1992,39, 315. ( 5 ) Mulvaney, P.; Henglein, A. J . Phys. Chem. 1990, 94. 4182. (6) Linnert, T.; Mulvaney, P.; Henglein, A.; Weller, H. J . Am. Chem. Soc. 1990, 112,4657. (7) Henglein, A. Chem. Phys. Lerr. 1989, 154, 473.
(8) Mostafavi. M.; Keghouche, N.; Delcourt, M. 0.;Belloni, J. Chem. Phys. Lett. 1990, 167, 193. (9) (a) Brkhignac, C.; Cahuzac, Ph.; Carlier, F.; Leygnier, J. Chem. Phys. Lerr. 1989,164,433. (b) Fallgren, H.;Martin, T. P.Chem. Phys. Lerr. 1990,168,233. (e) Wang, C. R. C.; Pollack, S.;Kappes. M. M. Chem. Phys. Lert. 1990, 166, 26. (d) Selby, K. Kresin, V.; Masui, J.; Vollmer, M.; Scheidemann, A,; Knight, W. D. Z , Phys. D 1991, 19, 43. (10) Harbich, W.; Fedrigo, S.; Buttet, J. Chem. Phys. Lerr. 1992, 195, 613. (1 1) Henglein, A.; Janata. E.; Fojtik, A. J . Phys. Chem. 1992,96,4734. (12) Ershov, B.G.;Sukhov, N. L.; Troitskii, D. A. Radiar. Phys. Chem..
in press.
(13) (a) Henglein, A.; Linnert, T.; Mulvaney, P. Eer. Eunsenges. Phys. Chem. 1990,94, 1449. (b) Henglein, A,; Mulvaney, P.; Linnert, T. J. Chem. SOC.,Faraday Discuss. 1991, 92, 31. (14) Henglein, A.; Tausch-Treml, R. J . Colloid Inferface Sci. 1981,80, 84.