The atmospheric sulfidation of silver in a tubular corrosion reactor

('orrosion Science, Vol. 29. No. 10, pp. 1179-1196, 1989

Printed in Great Britain

THE ATMOSPHERIC TUBULAR

0010-938X/89 $3.00 + 0.00 © 1989 Pergamon Press pie

SULFIDATION OF SILVER CORROSION REACTOR

IN A

L. VOLPE and P. J. PETERSON International Business Machines Corporation, General Products Division, San Jose, CA 95193, U.S.A.

Abstract---The corrosion of silver in air contaminated with sub-ppm concentrations of H2S was studied in a tubular reactor. Air with H:S or H2S-containing corrosive mixtures was passed through a circular tube lined with silver foil. Incorporation of corrodent into the metal sample was followed in real time by continuous monitoring of the inlet and outlet gas concentrations. After exposure, the silver sulfide formed on the foil was analysed. Applying a model for mass transfer in the tube, values of pollutant deposition velocity at the surface and of reaction probability ?, as a function of concentration, mixture composition, flow rate, reactor length, and exposure time were obtained. In dry air, HaS reacts with Ag at a constant rate corresponding to 7 on the order of 10 -~. The process probably involves 02 as a reactant. With the addition of nitrogen dioxide to the gas mixture, another reaction, 2Ag + HzS + 2NO 2 ~ Ag2S + 2HNO:, prevails. This reaction dramatically accelerates Ag2S formation. In a humid corrodent mixture with a high NO2/HES ratio, ), exceeds 10 -3, and the rate of sulfidation is controlled by the diffusion of HeS gas to the sample surface. Under most environmental conditions, mass transfer of pollutants in air is expected to be the main factor limiting the rate of silver corrosion.

C Cil

Ci C~;o~ D L L~ M Pe ppm r

R Rg Re (STP) T V Vavg X

),

NOMENCLATURE concentration of pollutant in the gas mixture concentration of H2S at the tube entrance concentration of HzS at the tube exit concentration of NO 2 at the tube entrance gas diffusion coefficient, diffusivity of HeS in air length of silver foil lining the reactor tube entrance length in the tube where the parabolic velocity profile is not fully developed molecular weight (of HeS ) P6clet number, ~;avgR/D parts per million by volume radial distance from the tube axis tube radius universal gas constant Reynolds number for air flow in the tube referred to standard temperature of 273 K and pressure of 1 atmosphere temperature air velocity average air velocity in the tube deposition velocity, ratio of H2S deposition rate to concentration of HeS at the surface distance down the tube from the leading edge of the foil reaction probability, probability that an HzS molecule colliding with the surface will form AgeS INTRODUCTION

metals in the atmosphere is decisively influenced by the type and content of pollutants in the air. 1-7 Among gaseous air contaminants, SO2, H2S, HC1, NO2, O3, and NH3 are commonly regarded as the most harmful. 5- 10 These reactive

T H E CORROSION o f

Manuscript received 23 April 1988; in amended form 15 November 1988. 1179

1180

L. VOLPE and P. J. PETERSON

gases are present in sub-ppm levels in ordinary environments. The corrosive action of the pollutant directly depends on its concentration at the metal surface. When a metal sample is exposed to corrosive air, the observed rate of corrosion is often limited as strongly by the delivery of corrodent to the sample as by the chemical reactivity of the metal. In other words, the observed rate is governed not only by the intrinsic reaction probability, i.e. the probability that a pollutant molecule colliding with the surface will react, but also by mass transfer of the pollutant in surrounding air. This is a typical example of series or mixed rate control. In flowing air under normal atmospheric conditions, the reaction probability 0') affects the rate of pollutant consumption when its value is <10 -2, for larger ~,values, gas-phase diffusion completely controls the rate. 11 When 9/is <10 -6 , the process is fully in the kinetic regime. These limiting values of Vdepend on flow conditions. The importance of gas-phase mass transport has been recognized in several corrosion studies. Haynie 12 gave a detailed analysis of the effect of air flow on the outdoor corrosion of zinc and galvanized steel. Finding that both the wind speed and sample geometry affected the rate of corrosion, he deduced a relationship between the air velocity and the corrosion rate. In a recent study of environmental variables affecting outdoor corrosion, Benarie and Lipfert 13connected the corrosion rate with the true delivery rate, or deposition velocity (Vd) of corrodent gases to the metal surface. The delivery rate was determined both by pollutant concentration and local wind velocity. In a laboratory investigation of the tarnishing of silver in the presence of H2S , C12, and SO2, Lorenzen 14 found that air velocity markedly affected the observed rate. Ishino et al. 15 studied in detail the effect of flow velocity on metal corrosion. Flat Ag and Cu plates were exposed inside a glass pipe to flowing air contaminated with H2S, SO2, and NO 2. A strong correlation between the tarnish rate and the gas-flow velocity was found. The kinetics of surface corrosion reactions can be separated from the details of corrodent delivery. It is important to study the intrinsic corrosion susceptibility of the metal independent of mass-transfer limitations. Since the diffusion of pollutants is usually slow at atmospheric pressure, it is difficult to study surface kinetics. Most of the reported corrosion rates, both in the field and in the laboratory, are partially masked by mass-transfer limitations. At low pressure, diffusion is faster but the corrosion reaction may be different. Vacuum conditions are unrealistic because they preclude a surface H20 film, a key condition for most corrosion processes. In this paper, a method allowing the determination of true surface kinetic constants from measurements of corrosion and pollutant deposition rates is introduced. The sample is a circular tube of metal foil, inside of which air with desired concentrations of corrosive gases is flowed. Corrodent concentration is monitored in real time at the tube inlet and outlet, giving the instantaneous rate of corrodent incorporation into the foil. After an exposure of the metal tube to the flowing corrosive air, the type and amount of solid corrosion product is determined. With the tube geometry, mass transfer of pollutant gas to surface is well-defined, and the deposition velocity and reaction probability can be easily derived. To evaluate the tubular reactor method, the tarnishing of silver by H2S was studied. Sulfidation is the major cause of Ag corrosion in polluted atmospheres. It has a high practical importance, especially in electronics. The reaction has been extensively studied in the field and the laboratory. 5'6'14'16-34 It is known that other widespread corrosive gases, notably NO2, can influence the rate of sulfidation with

Silver sulfidation in a tubular corrosion reactor

>3 cm

C°rr°sive gas mixture FIG. 1.

1181

~20 cm

l

]

Glass tube

Silver foil

El om d

Togas analyzer

Tubular corrosion reactor. The metal foil lines the tube inside diameter.

H2S. 5,29-31'34 Silver forms only one solid product, Ag2S, which makes its analysis relatively simple. EXPERIMENTAL METHOD The sulfidation reaction was studied by flowing air containing low levels of H2S through a silver tube (Fig. 1). The silver foil (Western Gold and Platinum Company, 99.9%, thickness 2.5 x 10-3 cm) was cold-rolled and had a surface roughness of 1 #m as measured with a stylus profilometer. It was washed sequentially in distilled water, acetone (distilled in glass), and hexane (distilled in glass). The foil was inserted tightly into a glass tube (borosilicate, length about 20 cm, ID 0.37 cm) so as to line it fully on the inside, except for a few centimeters near the flow entrance. Only the inner surface of the tube was exposed to reaction. In one type of experiment, shown in Fig. 2, the rate of Ag sulfidation in the cylindrical flow reactor was measured by comparing the inlet and outlet concentrations of HeS. At the inlet, sub-ppm concentrations of H2S or H2S and NO 2 were produced by flowing synthetic air (Liquid Carbonic, 21% 02, 79% N2, purified by passage through Drierite and zeolite traps at room temperature) past H2S and NO 2permeation tubes in a concentration generator (VICI Metronics, Dynacalibrator 450). The outlet concentration of H2S was monitored by converting H2S to SO 2 (Thermoelectron Corp. Model 340 Converter) and measuring the SO 2 concentration with a pulsed fluorescent SO2 analyser (Thermoelectron Corp. Model 43). Alternatively, the outlet flow was directed toward a chemiluminescenceNOx detector (Monitor Labs. Model 8840) to monitor NO and NO 2 levels. Both H2S/SO2 and NO wanalysers were calibrated by the permeation-tube method. The flow rate of air was controlled with a valve and measured with a mass flow meter. Flow rates of the order of 10 cm3(STP) s -l, corresponding to average flow velocities of 100 cm s 1 in the reactor tube were used. The reactor section was held at 293 K in a liquid constant temperature bath, and the reactor pressure was atmospheric. All wetted parts of the flow system, except for the Ag foil, were made of borosilicate glass, Teflon, and 316 stainless steel. This system was used to monitor H2S (or NO2) consumption by the silver sample continuously and to observe changes in the sulfidation rate with time and with changing process conditions. After an experiment, the silver corrosion product was examined visually and electrochemically. In another type of experiment, time-average sulfidation rates in the cylindrical reactor were measured. A manifold of five tubes was placed inside a large atmospheric-corrosion chamber (Fig. 3), similar to that described by Rice et al. 29 The chamber contained flowing air at 298 K, atmospheric pressure, and 70 +_ 3% relative humidity, contaminated with 0.50 + 0.02, 0.30 + 0.01, 0.040 _+ 0.005, and 0.0030 + 0.0005 ppm of NO 2, SO2, HzS, and HC1, respectively. The tube enfrance was open to the chamber. The chamber air

Constant-temperature bath Flow meter .............. and controller Ag tube ', "r

Purified synthetic air ~

i

G

~

'

c°ncentrati°n I from permeation tubes

Fic. 2.

.

x'/// .

.

.

.

.

.

.

.

.

.

.

NOx

analyzer

~

r~

Vent

i .

~

-Bypass

1 J

3 ~

~ H2S_to.SO2 converter

I

I ~ - Vent

SO2 analyzer

Apparatus for continuous monitoring of gas concentration.

1182

L. VOLPEand P. J. PETERSON

Flowing air 298 K 70% relative humidity 0.5 ppm NO 2 0.3 ppm SO 2 0.04 ppm H2S 0.003 ppm HCI

~

i

i

0

[]

--,"- i

I

0

[]

--~i

J

0

[]

~

!

!

0

~

I'

i

0 Silver filters

Silver tubes

•- Vent Pump

[]

[] Critical orifices

FI~. 3. Exposure of silver tubes to corrosion-chamber air. was pulled sequentially through the reactor tube and a membrane Ag filter35with a pump at a constant rate. The flow rate was controlled by a calibrated critical orifice at the exit of each tube. The flow rates ranged between 3.8 and 51.8 cm3(STP) s -l, corresponding to velocities of 38--518 cm s-1. The exposure times were 23.8, 24.0, 48.0, and 67.2 h. After the exposure the foil was removed from the tube and cut in sections along the reactor length. Each section was coulometrically reduced (electrochemically stripped) 36-38to determine the type and amount of corrosion product. This technique has long been known to identify reliably Ag2S, AgCI, and Ag20. The detection limit was on the order of 1/~g. The coulometric reduction yielded information on the corrosion product as a function of distance from the leading edge of the foil. DATA A N A L Y S I S Mass transfer of corrosive pollutant in air flowing t h r o u g h a cylindrical tube is described at steady state by

D{62C 1 6C + 7

62C~_ 6vC,

+ -LT

(1)

where C is the c o n c e n t r a t i o n and D a diffusion coefficient of the pollutant, r is the distance f r o m the tube axis, x the distance along the tube, and v the air velocity. The left- and right-hand sides represent diffusion and convection, respectively. T h e r e is no bulk reaction in the gas phase. In fully d e v e l o p e d laminar flow, the velocity distribution is V = 2Vavg 1

--

,

(2)

w h e r e Vavg is the average velocity and R is the pipe radius. With the flow rates e m p l o y e d , the R e y n o l d s n u m b e r (Re) is 41-370 for the real-time (concentrationmonitoring) experiments and 91-1235 for the time-average ( c o r r o s i o n - c h a m b e r ) experiments. These values c o r r e s p o n d to laminar flow. In the r e a c t o r entrance, the parabolic velocity profile is not fully developed. T h e entrance length for 99% d e v e l o p e d p a r a b o l a is L e = (1.18 + 0.112 R e ) R ,

(3)

and for 9 0 % - d e v e l o p e d profile Le is less than 1/2 of the 99% value. 39 Thus, for the real-time experiments, L~(99%) is 1.0-7.9 cm and for the time-average experiments 2.1-25.8 cm. Since the foil leading edge was located a few centimeters d o w n s t r e a m f r o m the tube entrance, the use of e q u a t i o n (2) is justified in analysing data f r o m the

Silver sulfidationin a tubular corrosionreactor

1183

real-time experiments. The parabolic velocity profile will be used to treat the time-average data, too, although at higher flow rates deviations from the model can be expected, especially near the leading edge. For all conditions, the velocity distribution is closer to the parabolic than to the flat, plug-flow profile. The relative importance of the axial-dispersion term D62C/6x 2, in equation (1) can be estimated. Under laminar conditions, D is the molecular diffusion coefficient of pollutant in air. It is calculated from the Chapman-Enskog kinetic t h e o r y ) ° At 293 and 298 K, D of H2S is equal to 0.158 and 0.163 cm 2 s -1, respectively, and D of NO2 is equal to 0.153 and 0.158 cm 2 s -1, respectively. The condition under which axial dispersion can be neglected was identified by Singh 41 for an analogous heat-transfer problem. In mass transfer, the contribution of the axial-dispersion term is negligible if the P6clet number, Pe = VavgR/D , is greater than 100. The D62C/6x 2 term is relatively small when the ratio is larger than 10. In the real-time experiments, Pe was 19-185, and in the time-average experiments the range of values was even higher. Thus, axial dispersion can be neglected. For laminar flow without axial dispersion, equation (1) is simplified to give

( r2) c

\dr 2 +r-~-r/

= 2V~vg 1 - - - ~

dx"

(4)

The equation is subject to boundary conditions C = Co 6C - 0 dr 6C Vd C - D 7 7 -r --

all r,

x = 0,

(5)

r=

allx,

(6)

all x,

(7)

0,

r = R,

where C0 is the entrance concentration, and Vj is the deposition velocity. In atmospheric science, Va is an experimental quantity determined by dividing the observed deposition rate by observed pollutant concentration close to, but not exactly at, the surface. 42'43 In equation (7), the boundary condition at the tube surface, the deposition velocity is simply a first-order or pseudo-first-order kinetic rate constant. From the gas-kinetic theory, the constant is equal to the reaction probability, 7, times the gas-surface collision frequency: Va =

/ R T \1/2 y 22~z_ 12ozM ) ,

(8)

where Rg is the universal gas constant, T the absolute temperature, and M the pollutant molecular weight. For reactions that are not first order, )' and Vd are functions of concentration at the surface. Here only pseudo-first-order kinetics will be considered. With the mixed control model, given by equations (4-7), the surface kinetic parameters lid and y can be derived from experimental data. Sideman et al. 44 provide an eigenvalue solution for an identical heat-transport problem. The data are compared with the solution for the mixing-cup concentration of pollutant, i.e. the mean concentration at a certain distance along the tube.

1184

L. VOLPEand P. J. PETERSON

EXPERIMENTAL RESULTS Real-time sulfidation rates Real-time experiments with H2S (and no NO2) were performed on two samples of silver foil. U n d e r constant inlet conditions at Co ~ 0.45 ppm and V,vg = 75 cm s -1, the conversion of the reactant gas decreased during the first 2 h and then came to a steady-state lasting as long as the process was continued, up to 16 h. When either the flow rate or the inlet concentration was varied, the conversion changed to a new steady-state value within 0.1-0.3 h. Table I lists the results of these experiments. The table also gives, for each set of process conditions, values of Vo and ~, calculated from the Sideman et al. solution to the reactor model. For Sample 1, the derived deposition velocities lie between 0.0092 and 0.051 cm s -1, corresponding to reaction probabilities between 0.86 and 4.80 × 10 -6. The values increase with increasing flow rate and with decreasing inlet concentration. The reactivities of the two samples are not very different under similar process conditions, indicating only a small sample-to-sample variation. In one run (Table 1), 3.8 vol.% H e was mixed with the inlet air. The addition of He had barely any effect on the rate. When a fresh Ag foil was exposed to air containing 0.5 ppm of NO2 (and no H2S), the outlet NO2 concentration first quickly fell by 0.1 ppm, but then started to rise, and in i h returned to the inlet value. In the next set of real-time measurements, H2S-NO2 mixtures were flowed through the tube. The rate of HeS consumption was monitored as a function of inlet NO2 concentration and foil length. As in the case of the H2S runs, each time the process conditions were varied, the rate came to a new steady-state value in 0.1-0.3 h, and then remained constant for many hours. Table 2 shows the results of these experiments. The derived values of Vo and ~, in Table 2 are markedly higher than those in Table 1. With Samples 3 and 4, the sulfidation rate increased orders of magnitude with increasing NO2 level. For Sample 5, the concentrations of nitrogen dioxide and hydrogen sulfide were held constant at the inlet, and the rate was monitored as a function of the foil length. With increasing distance from the leading edge, the calculated values of VO first

TABLE 1.

DEPOSITION OF

H2S FROM AIR ONTO Ag

TUBES

L

Vavg

Co*

CL*

Vd

Sample

(cm)

(cms -l)

(ppm)

(ppm)

(cms-l)

y x 10 6

1 1 1 1 1 1 1 1 1? 2 2

17.7 17.7 17.7 17.7 17.7 17.7 17.7 17.7 17.7 17.9 17.9

75.0 148 41.1 16.6 75.0 75.0 75.0 75.0 75.0 75.0 150

0.485 0.480 0.465 0.420 0.257 0.119 0.804 3.90 0.445 0.462 0.490

0.444 0.450 0.413 0.353 0.233 0.108 0.765 3.81 0.405 0.407 0.460

0.035 0.051 0.026 0.015 0.039 0.039 0.020 0.009 0.038 0.050 0.055

3.3 4.8 2.4 1.4 3.7 3.6 1.9 0.9 3.5 4.7 5.1

*All measurements at +0.03 ppm. "t3.8% H2 was added to air in this experiment.

1185

Silver sulfidation in a tubular corrosion reactor TABLE 2.

REACTIONOF Ag TUBES WITH FLOWING H z S - N O 2 MIXTURES

Sample

L (cm)

Vavg (cms -1)

CII* (ppm)

CL* (ppm)

CNo,* (ppm)

Vd (cms -1)

7 x 10 5

3 3 4 4 5 5 5 5 5 5 5 5

17.4 17.4 17.6 17.6 18.2 12.6 8.55 5.9 3.9 2.4 1.4 0.7

75.0 75.0 75.0 75.0 75.0 75.0 75.0 75.0 75.0 75.0 75.0 75.0

0.500 0.500 0.500 0.500 0.475 0.475 0.475 0.475 0.475 0.475 0.475 0.475

0.420 0,321 0.299 0.020 0.320 0.345 0.366 0.390 0.408 0.425 0.452 0.465

0.3(I 0.60 (I.60 3.44 0.62 0.62 0.62 0.62 0.62 0.62 0.62 0.62

0.072 0.19 0.23 4.9 0.16 0.19 0.24 0.26 0.31 0.37 0.27 0.2l

0.68 1.8 2.1 46 1.5 1.8 2.2 2.4 2.9 3.5 2.5 2.0

* All measurements are +_0.03 ppm.

increased from 0.21 to 0.37 and then decreased to 0.16. In Fig. 4, the conversion data are plotted, and the curve for Vd = 0.21 is given for comparison. When the reactor outlet was directed toward the N O x analyser, there appeared to be no drop in NO2 concentration and no N O production. This observation could not be accepted as a proof that N O 2 w a s not transformed in the sulfidation reaction because the instrument is k n o w n to suffer interference from HNO3, HNO2, and some other reactive nitrogen compounds, detecting them as nitrogen dioxide. 45 To check for the possibility of such an interference, the exit gas was passed through a bed

0.5

o ~

0.4 (/3 "6

o

g

0.3

E 8 oo

0.2

== 9

._x

0.1

0

5

10

15

20

Distance from leading edge (cm)

FIG. 4.

Reaction of H 2 S - N O 2 mixture with the silver tube wall. Temperature: 293 K. Air flow rate: 7.5 cm3(STP) s -l. Inlet NO2 concentration: 0.62 ppm. Curve: theoretical fit with assumed H2S reaction probability of 2 × 10-s. Error bar is smaller than data symbol.

1186

L. VOLPE and P . J . PETERSON

of Na2CO 3 powder upstream from the analyser. Sodium carbonate is known to trap nitric and nitrous acids, but not NO2, from dry air. 46With the Na2CO 3 filter, the NO 2 signal was substantially reduced. In particular, the conversion of NO2 was found to be approximately twice the conversion of H2S in these experiments. To identify the reactive nitrogen compound in the gas stream, a 1.0 ppm H2S + 0.3 ppm NO~ mixture was flowed through a 17.9 cm Ag reactor, so that all of the NO2 was converted. The outlet stream was bubbled through water. Then the aqueous solution was analysed by ion chromatography (Dionex 2120i chromatograph with an AS4A anion column). The only ion detected was NO2, and its amount was close to the amount of NO: converted during the run. The fact that nitrogen dioxide was quantitatively transformed into nitrite demonstrates that HNO2 is the gaseous product of the sulfidation reaction. Coulometric reduction of all the reacted foils (Samples 1-5) identified Ag2S as the only solid corrosion product. In an attempt to gain insight about surface reactive intermediates, transient tests were carried out. During a run with an HaS-NO2 mixture, the steady-state reaction rate was upset by a sudden cut-off of the NO2 or the H2S supply. When the NO2 was cut off, the conversion of H2S immediately decreased. The rate returned to the original value when NOz was reintroduced. Then the H2S was suddenly cut off; the conversion of NOz immediately went to zero. When the flow of H2S was restored, the rate of NO2 conversion was recovered at once. These results indicate that, under the reaction conditions, the surface is rich neither in reactive NOz species capable of oxidizing H2S nor in reactive H2S species that could reduce NO2 to nitrous acid.

Time-average sulfidation rates Figure 5 shows the sulfidation data of the time-average experiments with the gas mixture from the corrosion chamber. The data are presented as the thickness of Ag o

(,,)

I00-

801 6oi 401 2oi 0

150 "o =

lO( )o

Oo

5( "~9"~'°~ °'°'0 0

r~ "~'X'x-x--'+~'~'+'~'~"-'~'~-" 0

(b)

' '~ ....

<8

. . . . . 1~° . . . . .

.xx -++4, Do-o-O. o

~b -~

Distance from leading edge (cm)

20o

(c)

o ~-~-~

0 5 10 15 2O Distance from leading edge (cm)

250~D

(d)

50

1501~ o°

00 ~aa%0 "00

~X . ~ o ~ ' ~ - ~ " ° - ~ - o - o - --o-~

\'+ "~.a-,.~. 0 0 0 zr~'o 50 " t ~ . o .

x'~.+ O.o.o ~-A-.t,._ a %.

+'*-+-~.-~.-_~

- o-o

- o _ ~. o

0 , ~ , ,",~,.-r-..::, -,/T.~-~.--~,,-~+,, 5 10 15 20

Distance from leading edge (cm)

"~

5o~ '. + "°-o~_"S""~'~,-~."x,--~.+:+ ÷. ++, ~.-~3-o. D t + -D-o-o.~o 0 5 10 15

20 Distance from leading edge (cm)

FIG. 5. Sulfidation of silver foils with corrosion-chamber air. Data for flow rates (cm 3(STP) s-L): ©, 51.8; A , 27.3; rq, 14.3; +, 7.67; x , 3.83. Time of exposure (h): (a) 23.8, (b) 24.0, (c) 48.0, (d) 67.2.

FIG. 6. Photographs of Ag foils after 48 h exposure to corrosion-chamber air. Foil leading edge on the left. Flow rates through tube (cm3(STP) s-l): (a) 51.8, (b) 27.3, (c) 14.3, (d) 7.67, (e) 3.83.

1187

Silver sulfidation in a tubular corrosion reactor E

(a)

ls0F

~-

0

1189

(b)

o

o-//<.

10 20 30 40 50 60 70

~

Ld~-.--t',~"r-r'W--"*l, I ,

O0

Exposure time (h)

J

10 20 30 40 50 60 70 Exposure time (h)

/

80 r

,c, o/i

"0o I "o ~

•

40~-

~_

O0

10 20 30 40 50 60 70 Exposure time (h)

~-

O0

(d)

/°

/ o

10 20 30 40 50 60 70 Exposure time (h)

FIG. 7. Growth of Ag2S with time of exposure to corrosion-chamber mixture. Data for flow rates (cm3(STP) s-l): C), 51.8; A, 27.3; D, 14.3; + , 7.67; x , 3.83. Distance from leading edge (cm): (a) 1, (b) 5, (c) 10, (d) 18.

converted to Ag2S, averaged over each coulometrically reduced section of the foil, as a function of distance from the leading edge. With each of the four exposure intervals (Fig. 5a-d), the corrosion loss is given for five silver tubes (four tubes for the first interval) corresponding to different values of flow rate. The scatter of the points decreases with increasing length. The amount of sulfide formed markedly decreases with distance from the leading edge, as the H2S concentration in the flowing air is depleted. The rate of decrease is higher for smaller flow rates. The situation is vividly illustrated in Fig. 6, showing a photograph of the exposed foils. To determine the time-dependency of the sulfidation rate, the amount of Ag sulfided vs exposure time was plotted (Fig. 7). Figures 7a-d give interpolated data for distances of 1,5, 10, and 18 cm from the foil edge. Straight lines passing through the origin are drawn for all the data sets. The lines fit most of the data very well, which implies that the rate is constant with time. The largest deviations from linearity are for the shortest distance, but they do not seem to be systematic. Thus, the rate of Ag2S formation has no time-dependence for the present range of values of pollutant concentration, flow rate, foil length, and extent of sulfidation. In addition to the sulfide, coulometric reduction found AgC1 on some of the tubes, but the amount, corresponding to maximum AgCI thickness on the order of 1 nm was too small for quantitative analysis. Almost no chloride was formed on samples with higher flow rates and close to the leading edge, i.e. on foil sections having the largest amounts of sulfide. The chloride showed up in larger quantities downstream, where HzS was depleted, and where the sulfidation rate was low. To check if SO2 in the gas mixture was converted in the process, the SO2 analyser was attached downstream from one of the silver tubes. The concentration was equal

1190

L. VOLPEand P. J. PETERSON

to that in the corrosion chamber. This indicates that no SO2 was consumed during the sulfidation. DISCUSSION

Application of the model There is a range of process conditions where kinetic constants of the sulfidation reaction can be derived according to the tubular-reactor model. Figure 8 shows the theoretically expected decrease in HES concentration with distance down the Ag tube, for different values of reaction probability. With the flow rate of 7.5 cma(sTP) s -1 (Fig. 8a), the range of 7 values that can be deduced from the reactor conversion data lies between 10 -6 and 10 -3. Below 7 -- 10-6, the inlet and outlet concentrations are quite similar. Above 7 = 10-3, the mixing-cup concentration profiles are close to the profile for 7 = 1. The range of accessible 7 values can be shifted up or down by varying the air flow rate, total pressure, tube diameter, or tube length. Thus, with the flow rate of 83 cma(sTP) s -1 (Fig. 8b), one can extend the range up to 10 -2 at the expense of losing most of the resolution below 10 -5 . At both extremes of reactivity, the mixed control model is simplified. In the limit of low 7, the pollutant radial concentration profile is almost uniform, and the plug-flow equation disregarding mass transfer is applicable. In the limit of high 7 where the reaction is fast, the overall rate is completely governed by transport of corrodent to the metal surface. Gormley and Kennedy solved the equations for the case when the concentration of reactant gas at the tube wall approaches zero .47 Their mixing-cup concentration profiles coincide with that of Sideman et al. 44for 7 -- 1. The values of 7 obtained from real-time experiments (Tables i and 2) lie between the two extremes of reactivity, in the range suitable for the present kind of kinetic analysis. In dry air, the probability that a hydrogen sulfide molecule striking the corroding Ag surface will form silver sulfide is 10 -6 when H2S is the sole pollutant, and 10-5 when N O 2 is present at a level similar to that of HzS. Sulfidation with H2S in dry air The fact that without NO2 the calculated values of y vary with flow velocity and inlet H2S level indicates that the process deviates from the adopted mathematical model. There are three possible reasons for this deviation: a departure of the velocity profile from a perfect parabola near the leading edge, some degree of back-mixing or (a;

0 -.- 5 ~ ~ ~ O'3f~

(b)

",o -6 10 s

~ 0.40~

,o-, ~ O"1t l ~ " r ~ ~ ' ~ " ~ oo,

. . . . . . . . . . .

~. o . 5 0 ~ g"

~ v'vO 5 10 15 20 Distancefromleadingedge(cm)

O

~,.~

10-4-"

--

F

~ 0.30~. . . . . . . ,, .... , , , ~ O 5 10 15 20 Distancefrom leadingedge(cm)

FIG. 8. Theoretical consumptionof H2Sfrom air flowingin Ag tube for various assumed values of reaction probability. Tube diameter: 0.37 cm. Temperature: 293 K. Flow rates (cma(sTP) s-a): (a) 7.5, (b) 83. Values of reaction probability givenon the plots.

Silver sulfidationin a tubular corrosion reactor

t 191

axial dispersion, and surface kinetics other than pseudo-first order. The last possibility seems quite likely. The literature has little information about the dependence of the sulfidation rate on H2S concentration in air. Pope et al. 23 found that the rate was relatively insensitive to sulfide contents. Lorenzen 14 gave a correlation with a reaction order between 1/2 and 1. Crossland and Knight 25 estimated a 1/4-to-1/2 reaction order. Rice et al. 29 found the power of 0.14. Since those studies were performed under various degrees of mass-transfer limitations, the dependence of the rate on the surface concentration remains unresolved. To obtain the reaction order from tubular-reactor studies, one must modify the surface boundary condition (7) to include a power of C as the second adjustable parameter of the model. It is planned to use a two-parameter model in the future, after more data become available. Although much has been written about the reaction between Ag and H2S in air, the reaction mechanism is not fully understood, and even the stoichiometry remains controversial. In particular, it is unresolved whether the sulfidation proceeds directly, according to the overall reaction 2Ag + H2S = Ag2S + H 2,

(9)

or with the participation of atmospheric oxygen, according to the overall reaction 2Ag + HzS + ½0 2 = Ag2S + H20.

(10)

The controversy was discussed by Lilienfield and White 16as early as 1930. When those workers studied the reaction of Ag powder with H2S in moist air at room temperature, they found that silver sulfide was formed without evolution of H 2. They surmised that reaction (10) was taking place. With pure H2S, they observed H 2 production only at much higher temperatures. There is, indeed, no question that at high temperatures U2S sulfides silver directly, according to reaction (9). Saleh et al. 4s investigated adsorption of H2S on Ag film at temperatures above 193 K and found that at most about a monolayer could be adsorbed below room temperature. Bulk sulfidation and hydrogen-deuterium exchange began only above 423 K. Several groups of authors, 14"22'29following Lilienfield and White, speculated that reaction (10) was prevalent in air. Recently, however, Graedel et al., 32 studying the mechanism of atmospheric silver sulfidation, pointed out that the evidence for that reaction was not substantial and argued that reaction (9) was, in fact, taking place in the environment. To contribute to the understanding of this important reaction, H2 was added to the gas mixture in one of the runs (Table 1). At the process conditions, reaction (9) should be suppressed thermodynamically with less than 10 ppm of H 2. Yet, with a much higher hydrogen concentration, there was no decrease in the rate. This result argues against the direct sulfidation of H2S at room temperature. S u l f i d a t i o n with H2S-NO 2 m i x t u r e s

The fact that with HzS-NO 2 mixtures (Table 2) the rate of sulfidation increases dramatically with increasing NO2 levels agrees with the literature. Abbott 5 explored effects of NO2, SO2, and C12 on the rate of sulfidation by HzS in humid air. He observed that the added corrodents accelerated AgzS formation, in the order C12 > NO2 > SO2 > 02 - H 2 0 , and ascribed this order to the relative effectiveness of the gases in oxidizing HzS adsorbed on the surface to elemental sulfur. With NO2, he suggested

1192

L. VOLPEand P. J. PETERSON 8H2S + 4NO2 = $8 + 8H20 + 2N2

(11)

as a possible composite reaction. Workers from this laboratory29 reported no significant effect of NO2 concentration, but in this study this assessment was found to be incorrect. Studying the atmospheric corrosion of Ag by HzS , Guinement and Fiaud3° found a strong accelerating effect with NO 2 and a negligible influence with SO2. Since AgzS was the only solid product of sulfidation, they proposed that NO2 acted as a catalyst rather than a reactant. Hendriks 33looked at the reaction of Ag and Ag-Pd alloys with HzS in air and confirmed that the addition of other corrosive gases, especially NO2, has a strong synergistic effect on the process. The present study elucidates the synergism. In dry air at ordinary temperatures, the reaction between Ag and HzS without nitrogen dioxide is relatively slow, and NO 2 alone at the concentration used does not form a bulk corrosion product at an observable rate. With the mixture of the two gases, the 2:1 ratio of transformed NO 2 to HzS and the quantitative conversion of the reacted NO2 to gaseous ni:rous acid are consistent with the overall reaction 2Ag + H2S + 2NO 2 = AgES + 2HNO2.

(12)

Nitrogen dioxide serves as a reactant, not as a catalyst, and HNO 2gas is a product along with Ag2S. This process differs from that in the absence of NO2, and it is not surprising that it has a higher rate. In the atmosphere, NO 2 is usually one of the most abundant corrosive gas, more than an order of magnitude more abundant than H2S. 6-1°Therefore, reaction (12) should prevail as the major route of Ag tarnishing. Notably, since the solid sulfidation product is the same in all environments, the clue to observing this reaction lies in analysing both the gaseous reactants and products. The tubular reactor enables this to be done. The mechanism of reaction (12) is unclear. The transient experiments indicate that the surface does not have a high concentration of adsorbed HeS, NO2, or their mutually reactive intermediates at the process conditions. It is unknown whether other species, such as adsorbed S, have a large surface coverage. It would be interesting to find out if NO 2 oxidizes the surface first, and then H2S reacts with an oxide, or if adsorbed HzS and NO 2 react on the surface directly to produce S atoms. Also, surface H20, adsorbed as an impurity in the air stream, could influence the rate. In Fig. 4 for the NOz-HzSmixture, the theoretical curve for ?' = 2 × 10-5 exhibits a systematic deviation from the conversion data. The derived deposition velocity decreases with tube length. Such deviation is expected if the reaction order exceeds the value of 1 assumed by the present model. Since both reactant gases are depleted with increasing distance from the leading edge, the apparent reaction order must equal the product of the actual orders in HES and in NO2. This value is likely to be larger than 1, which may account for the observed deviation.

Sulfidation with the corrosive gas mixture from the chamber The mixing-cup concentration profiles of H2S in the Ag tube can be calculated from the measured rate of Ag2S formation on the wall, given the inlet or outlet concentration. The outlet concentration is known from the amount of Ag2S collected on the Ag filter downstream from the tube. Using the data for the longest exposure interval (Fig. 5d), we calculated the mixing-cup concentration, averaged over each foil section, starting from the outlet concentration and moving, section by section,

Silver sulfidation in a tubular corrosion reactor

1193

40 ~DoA 30

~,. ~ ' - o . "

20

\

o

+,,

\ ' x\~, +'\" \~

"'-

\[3.,

\ '\~g,' .. .

+,.

,q.. . . . . . . 5

-2"-_

~

" " Q~) , ....... .....

~...

v--

"~^

. ....

+-~-,.,. ~ - ~ .......... -C~, . . . .

o

~

~ +''+ .... +

"7' ~ "r- ~ ~ ~ ~ 10

-

~

"

+ ~

~ 15

[] "

-~-~

' 20

Distance from leading edge (cm)

FIG. 9. Depletion of H2S from corrosion-chamber mixture in Ag tubes. Experimental results for flow rates (cm3(STP) s-l): C), 51.8; A, 27.3; 13, 14.3; +, 7.67; x, 3.83. Uncertainty of concentration values is c a 5%. Prediction of model for reaction probabilities of 10-3 (dashed curves) and 1 (soLid curves).

toward the leading edge. The points in Fig. 9 represent the calculation. As expected, the concentration decreases with increasing length, and the rate of decrease is higher for lower flow rates. Along with the points in Fig. 9, theoretical curves are plotted for the same process conditions corresponding to sulfidation probabilities of 10 -3 and 1. All the experimental results lie in this range. The closeness of the curves for these two values of 7' does not allow the probability to be deduced more precisely. For three of the five values of flow rate, the points are closer to y -- 10 -3 near the tube entrance and closer to 7' = 1 near the exit. The sulfidation probability is much higher than that for the real-time experiments, and the conversion approaches the maximum value limited by gas-phase diffusion. There may be several causes for the higher effectiveness of Ag sulfidation with the corrosion-chamber mixture. First, the ratio of NO2 to HzS here is 12.5. In the real-time run with the ratio of 6.9 (Table 2), the reaction probability was already as high as 4.6 x 10 -4. The result is consistent with overall reaction (12) which, we believe, is dominant. Secondly, the concentration of hydrogen sulfide in the corrosion chamber is an order of magnitude smaller. If the reaction order in H2S is, in fact, less than 1, Vd should increase with decreasing concentration. Thirdly, the relative humidity in the time-average experiments was 70%, whereas the real-time experiments were done with dry air. Relative humidity is well known to have a positive effect on the rate o f s u l f i d a t i o n , 14J9'2°'23'25'3°--33although with gas mixtures similar to those employed the acceleration was found to be small. 29These three factors all could have contributed to the higher reaction probability. The influence of SO2 in gas mixtures is known to be relatively insignificant. 5'23'29"32The finding that SO2 does not react with the corroding surface is consistent with this fact.

1194

L. VOLPEand P. J. PETERSON

Although sulfide formation is the main process, a small amount of AgCl is also produced on the portion of the Ag foil where the rate of sulfidation is small or zero. U n d e r the process conditions, the reaction with H2S is much more favorable thermodynamically than the reaction with HCI. It is possible that the chlorination can begin only after the concentration of H2S near the surface goes to zero. Alternatively, the competition may be kinetic, i.e. AgCl is formed first but then is immediately displaced by the sulfide before it can build up in observable quantities. The sulfidation rate is time-independent even after a buildup of hundreds of Ag2S monolayers, in agreement with much of the literature.5'21'25'27'29"32This argues against the presence of solid-state diffusion limitations. The minimum value of sulfide thickness beyond which, at the process conditions, ionic diffusion may affect the rate, can be estimated. The electrical conductivity of the tarnish film at room temperature is 6 x 10 -4 ~-~-1 cm-1, 6% of it being due to Ag ÷ ions and the rest to electrons. 48-5° Then, from Wagner's theory, 51 the parabolic rate constant of film growth is 7 × 10 -11 equiv cm -1 s -1. The highest rate of sulfide formation achieved in these experiments, 1 cm from the leading edge of the foil, is 10 -7 cm s- 1. At this rate, diffusion constraints should be felt only after Ag2S thickness reaches several centimeters. Even if the rate were 6 n m s -1, corresponding to 7 = 1 at the inlet, the limiting thickness would have been several micrometers. Thus, the absence of any solid-state diffusion limitation is consistent with the high mobility of cations in Ag2S. In these experiments, the slowest of all processes is the delivery of H2S to the surface. Even at the highest air velocity, 5.3 m s -1, gas-phase mass transfer is the main factor governing the overall corrosion rate. In the environment, the same regime is expected to prevail. CONCLUSIONS (1) The tubular-reactor technique of passing corrosive air inside a cylindrical metal foil provides a useful method for controlled studies of atmospheric corrosion. The rate of corrodent incorporation into the solid was monitored in real time, and the time-average corrosion rate was obtained after the exposure. Mass transfer of corrosive gas to the surface can be calculated. One can derive the intrinsic gas-solid reactivity in the form of deposition velocity and reaction probability. In addition, gaseous products of corrosion can be analysed. (2) In dry air, H2S transforms silver into Ag2S with the probability of 10 -6. The rate is constant with time. The reaction does not go directly, with H Eevolution, which strongly suggests an involvement of 02 as a reactant. (3) With HES-NO2 mixtures in dry air, the reaction probability is 10 -5 when the concentrations of the two corrodents are similar. The rate rapidly increases with increasing NO2 level. Nitrogen dioxide reacts on the surface and is quantitatively converted to gaseous H N O 2 in the process. This accelerates Ag2S formation. This process is likely to be important in the environment because NO 2 is typically more abundant than H2S. (4) When a humid air mixture containing 0.5, 0.3, 0.04, and 0.003 ppm of NO2, SO2, H2S, and HCI, respectively, is used, Ag corrodes, forming the sulfide at a steady rate. The reaction probability with H2S lies between 10 -3 and 1. These high values are believed to result from the abundance of NO2, low H2S concentration, and high relative humidity. The sulfidation is so fast that even at high air velocities the observed rate is limited by gas-phase diffusion. The solid-state diffusion of Ag atoms from the bulk is faster than the interracial reaction and has no effect on the rate.

Silver sulfidation in a tubular corrosion reactor

1195

Under the process conditions, AgC1 is formed far less readily than Ag2S. In the atmosphere, mass transport of pollutant to the surface is expected to determine the rate of silver sulfidation. Acknowledgements--The authors are grateful to K. F. Sommers for experimental assistance in coulometric reduction and ion chromatography. A portion of this paper was originally presented at the Fall 1988 Meeting of the Electrochemical Society held in Chicago, Illinois. REFERENCES 1. H. R. UHLIGand R. W. REVIE,Corrosion and Corrosion Control (3rd edn), pp. 165-173. John Wiley, New York (1985). 2. E. MATTSON,Mater. Perform. 21, 9 (1982). 3. E. MATI'SON,Chemtech, April 1985, p. 234. 4. P. M. Aziz and H. P. GODDARD,Corrosion 15,529 (1959). 5. W. H. ABBOTT,IEEE Trans. Parts', Hybrids, and Packaging PHP-10, 24 (1974). 6. D. W. RICE, R. J. CAPPELL,P. B. P. PHIPPSand P. J. PETERSON,Atmospheric Corrosion (ed. W. H. AILOR), pp. 651--666. John Wiley, New York (1982). 7. D.W. RICE, R. J. CAPPELL,W. KINSOLVINGand J. J. LASKOWSKI,J. electrochem. Soc. 127,891 (1980). 8. R. ]~. COMIZZOLI,R. P. FRANKENTHAL,P. C, MILNERand J. D. SINCLAIR,Science 234,340 (1986). 9. T. E. GRAEDELand N. SCHWARTZ,Mater. Perform. 16, 17 (1977). 10. Environmental Conditions for Process Measurement and Control Systems: Airborne Contaminants, ISA Standard 71.04, Instrument Society of America, Research Triangle Park, North Carolina (1985). 11. J. A. GARDNER,L. R. WATSON,Y. G. ADEWUYI,P. DAV1DOVITS,M. S. ZAHNISER,D. R. WORSNOPand C. E. KOLB,J. Geophys. Res. 92, 10887 (1987). 12. F. I-t. HAYNIE,in Durability of Building Materials and Components, A S T M STP 691 (eds. P. J. SEREDA and G. G. LITVAN),pp. 157--175. American Society for Testing and Materials, Philadelphia (1980). 13. M. BENARIEand F. L. LIPFERT,Atmos. Environ. 20, 1947 (1986). 14. J. A. LORENZEN,Proc. 17th Annual Tech. Meet. Inst. Environ. Sei., Los Angeles, Calif., pp. 110-116 (1971). 15. M. ISHtNO, M. KISHIMOTO,K. MATSUIand S. MITANI, IEEE Trans. Comp. Hybrids Manuf. Technol. CHMT-3, 63 (1980). 16. S. LILIENFIELDand C. E. WHITE, J. Amer. Chem. Soc. 52,885 (1930). 17. L. E. PRICEand G. J. THOMAS,J. Inst. Metals 5,357 (1938). 18. H. FISCHMEISTERand J. DROTT,Acta Metal. 7,777 (1959). 19. J. DROaT, Ark. Kemi 15,181 (1960). 20. P. BACKLUND,B. FJELLSTRt)M,S. HAMMARB~.CKand B. MAIJGREN,Ark. Kemi 26,267 (1966). 21. R. W. CmARENZELLI,IEEE Trans. Parts, Materials, and Packaging PMP-3, 89 (1967). 22. W. E. CAMPBELLand U. B. THOMAS,Proc. 14th Holm Semin. on Electrical Contact Phenomena, pp. 233-265 (1968). 23. D. POPE, H. R. GIBBENSand R. L. Moss, Corros. Sci. 8,883 (1968). 24. H. E. BENNETT, R. L. PECK,D. K. BURGEand J. M. BENNETT,J. appL Phys. 40, 3351 (1969). 25. W. A. CROSSEANDand E. KNIGHT,Proc. 19th Annual Holm Semin. on Electrical Contact Phenomena, pp. 248-264 (1973). 26. K.-L. SCHIFF and H. BECKER, Proc. 9th Int. Conf. on Electrical Contact Phenomena, pp. 295-301 (1978). 27. S. P. SHARMA,J. electrochem. Soc. 125, 2005 (1978). 28. F. N. Fuss, R. F. LEACH and W. H. ¥OKOM, Proc. Int. Microelectron. Syrup. ISHM, Los Angeles, Calif., pp. 325-330. International Society of Hybrid Microelectronics, Montgomery, Alabama (1979). 29. D. W. RICE, P. J. PETERSON,E. B. RIGBY, P. B. P. PHtPPSand R. TREMOUREUX,J. electrochem. Soc. 128,275 (1981). 30. J. GUINEMENTand C. FIAUD, Proc. Int. Syrup. Testing and Failure Analysis, pp. 115-124 (1982). 31. W~ H. AaaoTT, Proc. Corrosion 83, Anaheim, Calif., Paper 234. NACE, Houston (1983). 32. J. P. FRANEY,G. W. KAMMLOTTand T. E. GRAEDEE,Corros. Sci. 25,133 (1985). 33. T. E. GRAEDEL,J. P. FRANEY,G, J. GUALTIERI,G. W. KAMMEOTTand D. L. MALM, Corros. Sci. 25, 1163 (1985). 34. A. H. C. HENDRIKS,Hybrid Circuits 11, 69 (1986).

1196

L. VOLPE and P. J. PETERSON

35. P.J. PETERSONand R. J. CAPPELL,Proc. Corrosion 83, Anaheim, Calif., Paper 235. NACE, Houston (1983). 36. L. E. PRICE and G. J. THOMAS, Trans. Electrochem. Soc. 76,329 (1939). 37. W. E. CAMPBELLand U. B. THOMAS, Trans. Electrochem. Soc. 76,303 (1939). 38. R. H. LAMBERTand D. J. TREVOY,J. electrochem. Soc. 105, 18 (1958). 39. B. ATICXNSON,M. P. BROCKLEBANK,C. C. H. CARD and J. M. SMITH,AIChEJ. 15,548 (1969). 40. J. O. HIRSCHFELDER, R. B. BIRD and E. L. SPOTZ, Chem. Revs. 44,205 (1949). 41. S. N. SINCH, Appl. Sci. Res. AT, 325 (1958). 42. R. LORENTZand C. E. MURPHY, Jr, Atmos. Environ. 19,797 (1985). 43. H. S. JUDEIKISand T. B. STEWART,Atmos. Environ. 10,769 (1985). 44. S. SIDEMAN,D. LUSS and R. E. PECK, Appl. Sci. Res. AI4, 157 (1965). 45. R. R. DICKERSON, A. C. DELANY and A. F. WARTBURG,Rev. Sci. Instrum. 55, 1995 (1984). 46. M. FERMand ~ . SJ()DIN, Atmos. Environ. 19,979 (1985). 47. P. G. GORMLEYand M. KENNEDY, Proc. Royal lrish Acad. 52A, 163 (1949). 48. H. REINHOLD and H. SEIDEL,Z. Elektrochem. 41,499 (1935). 49. G. BONNECAZE,A. LICHANOTand S. GROMB,J. Phys. Chem. Solids 39,299 (1978). 50. Yu. G. VLASOVand Yu. E. ERMOLENKO,Soviet Electrochemistry 17, 1068 (1981). 51. C. WAGNER,Z. Physikal. Chem. B21, 25 (1933).