Analysis of the Published Calorimetric Evidence forElectrochemical Fusion of Deuterium in PalladiumGORDON M. MISKELLY, MICHAEL J. HEBEN, AMIT KuMAR,REGINALD M. PENNER, MICHAEL J. SAILOR, NATHAN S. LEWIS*

Estimates are given of the raw data that are the basis for the claims of excess powerproduction by the electrochemical charging of palladium in deuterium oxide (D20).Calorimetric results are also presented that show no anomalous power production ineither 0.1M LiOD/D20 or 0.1M LiOH/H20 (LiOH is lithium hydroxide). Severalpossible sources of error in open-system calorimetry are discussed that can confoundinterpretation of temperature changes in terms of anomalous power production.

T HE PRIMARY OBSERVATION SUG-gested as evidence for the cold fusionof deuterium (D) in palladium (Pd)

is the production of excess power by the Pd,which was calculated to exceed the inputpower by factors of 4 to 8 under certainconditions (1). Numerous theories and ex-periments have been developed to explainthese effects, but the raw data that reflect theactual magnitude and conditions of the ob-served excess power production were notpresented in the original description of thework. The purpose of this report is topresent these raw data and an analysis of thecalculations used to obtain the large report-ed power production ratios. In addition, wedescribe some of our calorimetric measure-ments on the PdIO.lM LiOD-D20/Pt cell.We also describe several subtle experimentaldetails that must be considered to obtainaccurate enthalpy values for an electrolysisreaction in an open calorimetric system.When an aqueous solution is electrolyzed

to liberate hydrogen and oxygen gas, theelectrolysis power Papp (=Eapp I) can beconveniently partitioned into two terms(Fig. 1):

Papp = Eapp I = Pcell + Pgas (1)

Pgas represents the power removed as aresult of the evolution ofH2 (or D2 in D20)and 02 gases, and Pcell is the remainingpower that is effective in heating the cellcontents. An expression for Pgas (= Egas I) isreadily obtained from the known enthalpyof formation of water from its elements:Egas = - AHform/2F (F is Faraday's con-stant), which yields Egas = 1.48 V for thereaction H20 -- H2 + 1/202 and Egas =

1.54 V for the reaction D20 -* D2 + '/202(2). If the net faradaic efficiency of gasevolution, K, is known, Eq. 1 for D20electrolysis then becomes

Papp = Eapp I = [(Eapp - K 1.54 V) I]+ K [(1.54 V) I] (2)

Division of Chemistry and Chemical Engineering, Cali-fornia Institute of Technologv, Pasadena, CA 91125.

*To whom correspondence should be addressed.

10 NOVEMBER 1989

with the first term on the right equal to Pcelland the second term equal to Pgas. If no gasescapes from the cell (because of efficientrecombination processes), then K = 0, andEq. 2 yields the expected Papp = Pcell, withall of the applied power resulting in heatingof the cell contents. Alternatively, if thegases escape with unit efficiency, thenK = 1, and Pceel = (Eapp - 1.54 V) I. Thislatter condition results in the minimumheating of the cell for a given value of Papp.

In their analysis of the calorimetric data(1), Fleischmann, Pons, and Hawkins as-sumed K = 1 and therefore defined the ex-cess power, Pex, as the difference betweenthe measured heating power in the cell,Pmeas, and the expected value of Pcell= (Eapp - 1.54 V) I. This definition, com-bined with reported values of the excesspower (1) and the various formulas used tocalculate "percent of breakeven values" (seebelow), has allowed us to recalculate the rawcalorimetric data for the experiments ofFleischmann et al. (1), which are presentedin Table 1 (3).

Several features of the data in Table 1 areworthy of discussion. First, in most cases,the calculated values of PeX are rather smallpercentages of the total input powers. Accu-rate error bars for the determination of Papp,

=k

TbTher

Fig. 1. Schematic ofthe calorimeterused in this work. The power ap-plied through the electrolysis cir-cuit Papp==Eapp Iapp producesheating power delivered to the cell,Pcell, and to power contained in theevolved gases, Pgas. The tempera-ture difference between the electrol-ysis cell and the constant-tempera-ture bath is used to determine theheating power of the electrolysiscircuit. Calibration and constancyof the heating coefficient, as well asthe presence of thermal gradients,are key issues of concern for accu-rate power measurements in thissystem.

Pgas and Pmeas must be obtained to assessthe significance of the reported Pex values.Second, recombination losses have been as-sumed to be negligible (that is, K = 1.0) incalculating P,,, and P,X; if substantial re-combination occurred, Pe,,e would increase,and this would result in smaller, or possiblynegligible, values of PeX. Only in the last twomeasurements does Pmeas (= Pex + Pcell)exceed Papp; thus, only in these two casescould an adjustment in the value of K notcompletely account for the observed data.The extent ofrecombination ofH2 and 02 iSa sensitive function of the cell geometry,electrode history, voltage, use of a separator,and so on (4, 5), and, although it may beunlikely that K approaches zero for all thecells under study, a rigorous analysis of thecalorimetric data requires that K be reportedin order to obtain reliable values for Pcelland Pex. Possible solvent evaporation lossesmake mass balance determinations based onthe rate of liquid D20 loss (1) a question-able method for establishing the value of K;a more reliable method would be collectionof the gases followed by volumetric andcompositional analysis during the calorimet-ric measurements.

It is also of interest to outline the proce-dures used by Fleischmann et al. (1) tocalculate the percent of breakeven valuesfrom the raw data in Table 1. In all cases, Pexwas defined as Pex = Pmeas - Pce, assumingK = 1.0 [that is, Pcell = (Eapp -1.54 V) I].The first calculated breakeven percentagevalue [column a in table 2 of (1)] thenrepresents (100 X Pex)/Pcell; column b intable 2 of (1) represents (100 x Pex)/Papp.In both cases, decreases in K from the as-sumed value of 1.0 would reduce the calcu-lated values. In calculating the raw data ofTable 1 above, we have used the percentbreakeven data in column a in table 2 of (1),and the formula (100 X Pex)/Pcell to calcu-

REPORTS 793

on

June

4, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

late P.,, (and thus Eapp); in some cases,using the reported data in column b in table2 of (1), with the formula (100 x Pex)/Pappyields substantially different values for Eapp,and these values have been indicated inparentheses in Table 1. In general, the Pex'Papp ratios are rather modest and need to beconsidered carefully in view of possiblechanges in K and errors in the determinationof Pmeas.The actual applied voltages range from

2.8 to 10 V (Table 1), but column c in table2 of (1) uses a voltage of 0.5 V to obtainPapp = [(0.5 V) I]. The percent breakevenvalues in this column have been calculated as(100 x Pe,)/[(0.5 V) I]. Of course, theassumption ofE = 0.5 V yields much high-er breakeven values than are calculated fromthe measured data, which display much larg-er Eapp and Papp values.The choice of a hypothetical value of 0.5

V involves the scenario where the electroly-sis reaction does not evolve 02 at the anodebut rather oxidizes the D2 generated at thecathode back to D20. In this case, only thevoltages (overpotentials) required to over-come the electrode polarizations, solutionresistance, and concentration gradientswould contribute to the cell voltage, as theelectrolysis voltage necessary to produce D2and 02 from D20 is eliminated. In princi-ple, in optimal cell design, it is possible toenvision a reduction in voltage to <1 Vbetween two electrodes that have no otherovervoltages. However, if the anode is tooxidize D2, it must operate at potentials -O

Fig. 2. Plot of the potentiostatic cur- 0rent-voltage characteristics for acharged Pd/D rod (0.254 mm in Ediameter, 2.7 cm long, charged for14 hours at 64 mA/cm2) in 0. 1M E Go -LiOD/D20. Potentials are with re- ,spect to a Pd wire charged to the a-0 nphase equilibrium; the reference wire *was in a separate compartment filled Xwith 0.1M LiOD/D20. This refer- 0ence had a potential of -0.952 V ,with respect to a conventional satu-rated calomel electrode. (-) Directlymeasured data, scan rate = 10 mV/s; -120 -(-* ) data corrected for solutionresistance; (O) potential measured 5ms after interruption of a current density of 75 mA

V versus a reference Pt/D2 electrode, whilethe instantaneous voltage measured afterinterruption of a current of 75 mA/cm2shows that the potential of a Pd/D electrodeis 0.7 V more negative than that of a PdlDreference charged to the ot-I phase equilibri-um (Fig. 2) and becomes even more nega-tive when current is flowing through thecircuit. Thus, the electrochemical potentialrequired to maintain reasonable currentdensities through the charged Pd cathode isgreater than the value of 0.5 V assumed byFleischmann et al., implying that the largehypothetical breakeven values calculatedwith the use of a value of 0.5 V are notexperimentally attainable.The measured quantity in the calorimetric

experiments of Fleischmann et al. (1) is a cellpower, not an enthalpy of reaction. Mea-surement of the power produced requires an

Table 1. "Raw data" for electrolysis ofD20 at Pd. The data were calculated from table 1 and column ain table 2 of (1). The expected power produced in the cell was calculated from (E - 1.54) I, that is, onthe assumption that there was no recombination ofD2 with 02. The excess power produced in the cell isas quoted in table 1 of (1). The values in parentheses were calculated from column b in table 2 of (1). Inall other cases, values calculated from column b agreed with those calculated from column a in table 2 of(1). The highest current entries for all three rods were measured on rods 1.25 cm long and rescaled tocorrespond to rods 10 cm long. In these calculations, we have ignored the areas at the end of the rodsand have rounded the applied voltages to three significant figures while attempting to preserve thesignificance of the excess power data quoted in (1).

Applied Applied Input Power Expected Excess

current, voltage, power, produced power powerI E Papp = E I in cell, produced produced

(mA) (V) app- Pmeas in cell in cell(mA) (V) (W) ~~~ ~ ~~~~~(W)(W) (W)

Rod 0.1 cm in diameter and 10 cm long25.13 2.84 0.0714 0.0402 0.0327 0.0075

(2.49) (0.0626) (0.0314) (0.0239)201.1 3.61 0.726 0.495 0.416 0.0791608 9.67 15.55 13.73 13.07 0.654

(8.13) (13.07) (11.25) (10.60)Rod 0.2 cm in diameter and 10 cm long

50.27 2.70 0.136 0.094 0.058 0.036402.1 4.21 1.696 1.57 1.074 0.4933217 8.25 26.5 24.6 21.59 3.02

(8.54) (27.5) (25.5) (22.52)Rod 0.4 cm in diameter and 10 cm long

100.53 2.91 0.293 0.291 0.138 0.153804.2 4.84 3.89 4.40 2.654 1.751

6434 8.60 55.3 72.2 45.4 26.8

794

1cm2.

-1.5 -1.0 -0.5Potential (Vversus Pd/D)

0.0

accurate calibration of the rate of heat lossfor all conditions of interest. To addressthese issues, we have performed several calo-rimetric experiments with both 0.1MLiOD/D20 and 0.1M LiOH/H20, usingeither a commercial isoperibolic calorimetrycell (Tronac, 50 cm3) in a 60-liter Tronacthermostatted bath (27.000° ± 0.005°C) oran air-jacketed vessel (30 cm3) in a 5-literVWR bath (25.000 ± 0.040C). Both cellswere mechanically stirred and were calibrat-ed throughout the experiments using theheat output from 100- to 340-ohm resistors.Current for the electrolysis and the calibra-tion was supplied from Princeton AppliedResearch potentiostats (models 173, 273, or362) or from Hewlett-Packard power sup-plies (model 6002A). Currents and voltageswere measured with Fluke 75 or Keithley177 multimeters, and temperatures weremeasured with the use of thermocouples,with the reference at the temperature of thesurrounding bath. The counter electrodeswere cylinders of platinum (Pt) foil or mesh(2 cm in diameter, 2 cm long), and the Pdelectrodes were 2.0- to 2.4-cm lengths ofrod of diameters 2.2, 2.1 to 2.3, or 3.9 mm,contacted either by spot welding to Pd or Ptwire, or by soldering to copper wire. Allelectrodes were completely immersed in thesolutions at all times, and connections wereinsulated from the solution with glass tub-ing and epoxy resin. The electrolyzed waterwas typically replaced daily, during whichtime <5% of the solution was electrolyzed.These calorimetric measurements led to

the following observations. First, at all cur-rent densities (8 to 150 mA/cm2) and times(up to 170 hours) the measured heatingpowers in the LiOD/D20 cells were within5% of the expected power calculated fromEq. 2 if we assume K = 1.0, and theLiOH/H20 experiments gave results towithin 6% of the predicted values. [Heatlosses due to evaporation at the highestoperating temperatures (49°C) are calculat-ed to cause errors of less than 2% in thecalculated heating power.] These measure-

SCIENCE, VOL. 246

on

June

4, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

ments were made by two methods. In initialexperiments the current density waschanged, and the power applied through acalibration resistor was adjusted to maintaina constant cell temperature (Table 2). Forcomparison, the reported excess specificpower production for a Pd rod 0.20 cm indiameter (1) implies that an excess power of0.14W would be expected for our specimenat a current density of 64 mA/cm2, whereasan excess specific power of 10 W/cm3 wouldcorrespond to an excess power of 0.91 Wfor our sample. In later experiments theresistor power was stepped at constant elec-trolysis current density, and the resultingplot of the total applied power versus tem-perature was used to determine the heatingpower from the electrolysis circuit.

Several runs with the first method showedchanges in the cell temperature that did notcorrespond to adjustments in the appliedcurrents or voltages. These changes wereclearly caused by changes in the heat transferfrom the cells, rather than by some addition-al heating mechanism, because recalibrationshowed that the input electrolytic heatingpower was equal to the calculated enthalpy-corrected electrolysis power (Pceii withK = 1.0) both before and after the tempera-ture changes. The temperature changes typi-cally were slow increases (often startingabout 20 hours after the start of charging),and more rapid decreases, which occurredeither spontaneously or occasionally afterthe addition of D20 or H20. Withoutrecalibration of the cell, such temperatureincreases could have been interpreted interms of increased power production, butthey actually reflected decreases in the rate ofheat loss from the cell to the surroundings.To differentiate further between an in-

crease in output power and a change in rateof heat transfer, we measured the heatingcoefficient (the ratio of the temperature riseto the input heating power) of the Pdelectrolysis cell by holding the electrolysiscurrent density constant and incrementingthe power applied through the calibrationresistor. Figure 3 shows the dependence ofthe cell temperature upon the total appliedheating power for a 0.1M LiOH/H20 cell,on the assumption that heating is the resultof the sum of the calibration resistor powerand the enthalpy-corrected electrolysis pow-er (Pceii with K = 1.0) of the Pd/Pt circuit.At a particular Pd/Pt electrolysis currentdensity, the temperature at the location ofthe thermocouple was linearly dependent onthe total input heating power over the tem-perature range 250 to 40°C, with a slopeequal to the heating coefficient of the cell.Initially the heating coefficient decreasedwith increasing current density, consistentwith additional heat losses caused by gas

IO NOVEMBER I989

evolution, but 23 hours after the start ofcharging in the cell with a rod 3.9 mm indiameter the temperature increased. If thetemperature rise had been due to an addi-tional source of heat, the new heating linewould have been parallel to the initial line,but displaced upward. Instead, the subse-quent points defined a steeper line, indicat-ing that the heating coefficient changedfrom 120 to 14°C per watt. Analysis of thedata on cell temperature versus time indicat-ed that the actual heating power producedby the electrolytic circuit remained constantto within 6%. Changes in heat transferduring bubble evolution in open systemshave been reported (6, 7), and it is possiblethat a similar effect is occurring in thepresent system; in addition, changes in thevolume of liquid owing to gas evolution andevaporation can also contribute to a time-dependent change in the heating coefficientof the cell (8), as could changes in thethermal conductivity of the cell walls or theelectrodes (9). Moreover, at higher operat-ing temperatures of the Dewar flask, nonlin-earity in the relation of temperature to heat-

ing power of the calorimeter can lead tosubstantial overestimates of the electrolysisheating power, as can improper positioningof the temperature-sensing element relativeto the resistive heater (10, 11). Clearly,accurate calorimetric measurements in thistype of open cell require efficient stirring ofthe solution to eliminate temperature gradi-ents, quantitative collection of the gasesevolved, and calibration measurements at allelectrolysis current densities and after alltemperature changes; otherwise, it is notpossible to unambiguously assign a rise incell temperature to an increase in powerproduction by the electrolysis circuit.

In conclusion, analysis of the data report-ed earlier (1) shows that the observed powerproduction is much smaller than generallyreported in the popular press and that sever-al assumptions are required to transform theraw data into large excess power productionratios. In other recent experiments for whichresearchers have claimed excess power pro-duction in Pd/D2O/Pt electrolysis cells time-dependent changes have been observed inthe slopes of electrolysis power versus tem-

Table 2. Representative calorimetry data for H20 and D20 electrolysis cells. The total power wascalculated as the resistor power plus the electrolysis power. The measured temperatures are +0.04°C;bath temperature = 27.000° ± 0.005°C. The error bars quoted were based on 2a values for randomerrors in the multimeters and temperature measurement devices. The comparison between H20 andD20 gives an estimate of the magnitude of any systematic errors in the calorimeter.

Time Current Electrolysis Resistor Total Temperature(hours) density power power power C( (mAJcm2) (W) (W) (W) 'O'

Drawn and machitned Pd rod (0.21 to 0.22 by 2.1 cmn), 0. 1M LiOH/H2034.5 59 0.174 ± 0.002* 0.118 ± 0.001 0.292 ± 0.003 42.6246.5 86 0.292 ± 0.003* 0 0.292 ± 0.003 42.6867.5 59 0.175 ± 0.002* 0.119 ± 0.001 0.294 ± 0.003 39.8277.5 86 0.297 ± 0.003* 0 0.297 0.003 39.92

Drawni Pd rod (0.22 by 2.4 cmn), 0.1M LiOD/D2044.0 66 0.253 ± 0.003t 0.250 ± 0.002 0.502 ± 0.005 49.3646.0 97 0.457 ± 0.003t 0.044 0.501 ± 0.003 49.3972.0 97 0.473 ± 0.003t 0 0.473 ± 0.003 45.9990.0 66 0.277 ± 0.003t 0.204 ± 0.002 0.481 ± 0.005 46.01

*Elcctrolvsis power = (E - 1.48 V) I. tElectrolsis power = (E - 1.54 V) I.

Fig. 3. Plot of the observed 5temperature versus the ap- 36'plied heating power for aPd/H rod 3.9 mm in diame-ter and 2.4 cm long in 0.1M 0

LiOH/H20, on the assump- @ 32tion that there is no recom- I./bination of evolved gases /[that is, heating power = I E(E - 1.48 V) + Presistor]. 28First, the calibration resistorwas stepped in power withno electrolysis (0). Thencurrent was applied at 8 24m.A/cm2 for 17.1 hours, 0.0 0.2 0.4 0.6 0.8 1.0during which time the cali- Applied heating power (W)bration resistor power was again stepped (A). The electrolysis current was then increased to 60 mA/cm2for the remainder of the experiment (0). The labeled points were obtained at the following times afterthe second current step: 1, 3.6 hours; 2, 5.8 hours; 3, 8.7 hours; 4, 17.0 hours; 5, 21.2 hours; 6, 26.1hours; 7, 30.4 hours. After 12 hours at 60 mA/cm2, 0.7 cm3 of H20 was added to the cell (total cellvolume = 30 cm3).

REPORTS 795

on

June

4, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

perature plots (12), as opposed to changes inthe intercepts of such plots. These data areconsistent with a change in the heat lossconstant (Fig. 3), as opposed to an increasein the rate of power production by theelectrolysis cell. Careful calibration of therate of heat loss and accounting for allsources of additional heat generation (gasrecombination and so forth) are requiredbefore one can unambiguously assign achange in temperature of the electrolysis cellto an unexplained chemical or nuclear pow-er-producing process. If required, more ac-curate power production measurements onthe Pd/D20/Pt system could only readily bemade in closed-system calorimeters, where awell-defined heat path can be establishedand total recombination of the gases can beassured.

Note added in proof In recent experimentsin a closed-system calorimeter with totalrecombination of the gases, a recast Pd rod(0.39 cm in diameter and 1.0 cm in length)at 58 mA/cm2 current density in 0.1MLiOD/D20 produced no measurable excesspower [Pmesa = (100 + 2%) of Papp] for aperiod of over 12 days.

C LOSELY PACKED GRANULAR SOLIDSwith interstitial pore spaces filled byliquid occur in both natural and

man-made environments. Familiar examplesinclude water-saturated soil and fragmentedrock that compose most landslides, debrisavalanches, and debris flows. Other exam-ples include industrial slurries and granularmixtures, as well as saturated sediments thatmay be sheared during tectonic faulting.

Observations of mass-movement process-es on the earth's surface have motivatedfundamental questions about the dynamicrole of pore water in rapidly sheared granu-lar materials. For example, how can a water-

REFERENCES AND NOTES

1. M. Fleischmann, S. Pons, M. Hawkins, J. Elec-troanal. Chem. 261, 301 (1989).

2. S. Gross, Energy Cotivers. 9, 55 (1969).3. A similar calculation has been independently per-

formed: G. Kreysa, G. Marx, W. Plieth, J. Elec-troanal. Chem. 266, 437 (1989). There are twoerrors in column c in table 2 of (1). The values forthe rods of 0.2- and 0.4-cm diameter at 8 mA/cm2should be 143 and 306%, respectively.

4. J. M. Sherfey and A. Brenner, J. Electrochem. Soc.105, 665 (1958).

5. M. J. Joncich and N. Hackerman, J. Phys. Chem.57, 674 (1953).

6. R. G. Edkie and P. L. Khare, Int. J. Heat MassTransfer 15, 261 (1972).

7. S. C. Bhand et al., ibid. 8, 111 (1965).8. W. E. Meyerhof, D. L. Huestis, D. C. Lorents, in

preparation.9. A. W. Szafranski and B. Baranowski, J. Phys. E 8,

823 (1975).10. R. Garwin, Natu4re 338, 616 (1989).11. N. S. Lewis et al., paper presented at the Workshop

on Cold Fusion Phenomena, 24 May 1989, SantaFe, NM.

12. R. A. Huggins, ibid.13. Supported by the Office of Naval Research and

under a special grant from the California Institute ofTechnology. We thank G. Kreysa and W. Meyerhoffor supplying preprints ofrelated work. We thank C.A. Barnes and S. E. Kellogg for helpful criticism ofthis manuscript. Contribution 7969 from the Divi-sion of Chemistry and Chemical Engineering at theCalifornia Institute of Technology.

21 June 1989; accepted 29 August 1989

saturated mass of landslide debris liquefyand flow tens of kilometers across slopes asgentle as a few degrees-much farther thansimilar masses of dry debris and much far-ther than anticipated from the material'sinitial potential energy and coefficient ofinternal friction (1)? In most theories, lique-faction caused by high, quasi-static pore-water pressure gradients has been invokedto account for the mobility of wet rockdebris (2), but high pore-pressure gradientsare unlikely to persist in debris that flowssteadily for minutes or even hours. Further-more, water is an ineffective lubricant onmost rock surfaces (3). Thus, pore watermust enhance efficient, rapid shear deforma-tion of rocky debris by some alternativemeans.

In this paper we describe experiments

Fig. 1. Cross-sectional view through the center ofthe array of fiberglass rods. Water pressures weremeasured in individual pores indicated by num-bered dots. Arrows show relative motion alongthe prescribed slip surface.

testing a new hypothesis: that dynamicpore-pressure fluctuations can be generatedas a result of grain rearrangements duringrapid shear, and that the fluctuations can belarge enough to modify grain-contact stress-es significantly and promote efficient defor-mation. Specifically, transient increases inpore pressure in discrete domains where thegranular phase momentarily contractswould inhibit further contraction and easelocal shear displacement. Conversely, wherethe granular phase dilates, transient reduc-tion in pore pressure would suppress furtherdilation and inhibit local shear displacement.In a deformation field in which local dila-tions and contractions accompany globalshear, pore-pressure fluctuations may reducefrictional energy dissipation because theyhelp localize intergranular shearing in areasof low or zero grain-contact stress (4). Sucheffects are enhanced if the pore fluid (forexample, water) is relatively incompressible.An important lemma is that local dilationand contraction can occur even while thebulk deformation is steady (5, 6).The tendency for pore-pressure fluctua-

tions to develop during steady shear defor-mation depends on the relative rates of grainrearrangement and pore-pressure equilibra-tion, which can be expressed by a singledimensionless parameter, R = kElvp8 (7, 8)where k is the hydraulic permeability, E isthe uniaxial (Young's) compression modu-lus of the composite granular medium, v isthe velocity of intergranular sliding, uti is theviscosity of the pore fluid, and 8 is a charac-teristic length, typically the grain diameter.Within R the quantity kEIl, functions as apore-pressure diffusivity, and 82pJIkE is thetime scale for diffusive pore-pressure equili-bration over the distance 8. The quantity b/vis the time scale for pore dilation and con-traction and thus for generation of pore-

SCIENCE, VOL. 246

Dynamic Pore-Pressure Fluctuations in RapidlyShearing Granular Materials

RICHARD M. IVERSON AND RicHARD G. LAHUSEN

Results from two types of experiments show that intergranular pore pressuresfluctuated dynamically during rapid, steady shear deformation of water-saturatedgranular materials. During some fluctuations, the pore water locally supported allnormal and shear stresses, while grain-contact stresses transiently fell to zero.Fluctuations also propagated outward from the shear zone; this process modifiesgrain-contact stresses in adjacent areas and potentially instigates shear-zone growth.

U.S. Geological Survey, Cascades Volcano Observatory,5400 MacArthur Boulevard, Vancouver, WA 97661.

796

on

June

4, 2

009

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om