Tables 2.1 and 2.2 summarize many experimental efforts aimed at verification of the excess heats originally reported by Pons and Fleischmann. The Panel is aware of other efforts that have remained unreported in any formal way. It was not possible to visit all of the active laboratories, but the Panel did send representatives to the Pons-Fleischmann laboratory and the Wadsworth laboratory at the University of Utah, the Huggins laboratory at Stanford University, the Lewis laboratory at Caltech, the Appleby, Martin, and Bockris laboratories at Texas A&M, and to McKubre of SRI at EPRI. In addition, detailed queries were made of other investigators by telephone or by conversation at meetings. It is not practical here to provide a detailed analysis of each group's work. Instead, we will provide a summary of the kinds of experimental problems that we identified in the calorimetric measurements.
One of the important ideas advanced in this field is that large amounts of heat arise from a previously unrecognized source, hence it was the one of the Panel's chief responsibilities to evaluate experimental work from the positive claimants with great care. The experimental shortcomings discussed here were widely encountered. No single study was compromised by all of them, but no positive report that we were able to study in detail was free of such problems.
The experimental schemes used in most laboratories reporting positive heat effects normally have many elements in common. A cell containing a Pd cathode and a Pt anode immersed in D2O (or H2O) containing LiOD (or LiOH) is allowed to pass a steady current for a long period of time. For those which have followed the Fleischmann-Pons method (ISO A), the cell body is immersed in a heat sink (usually a water bath), and precise measurements are made of the temperature difference between the interior of the cell and the heat sink. Thermocouples and microvoltmeters are usually used for the temperature measurements, but sometimes thermistors are employed. The evolved heat power is usually calculated from the temperature differential by a heat transfer coefficient, which is obtained by calibration via the input of additional heat power at a resistor immersed in the working solution. The variant of this scheme to make it more truly isothermal (ISO B), and avoid possible temperature dependence of the heat transfer coefficient by standard heat substitution, combines calibration and measurement.
An increasing number of laboratories are now reporting results with calorimeters that measure heat flux directly (ISO C) but calibrations with power dissipated through standard resistors in the cell compartment are still required. This is also true for flow calorimeters.
The claims of excess heat rest upon a comparison between the measured heat power evolved from the cell and the electrical power input. The latter quantity is determinable as the product of the current through the cell and the voltage across it. Comparisons normally involve a correction of this quantity by subtraction of 1.527 V times the cell current, to account for the enthalpy leaving the cell as vented D2 and O2. The underlying assumption is that 100% of the current goes to produce D2 and O2, which is vented perfectly. It is possible to operate the cell in a closed fashion by including a catalytic recombiner capable of regenerating deuterium oxide from D2 and O2. Very rarely has excess heat been reported from such a cell.
In most cases, the current through the cell is held constant, and the voltage is allowed to fluctuate according to variations in mass transfer rates and electrode surface exposure in the cell. (These occur largely because of the intense gas evolution normally taking place in the cells.) In other experiments, the total applied voltage is controlled, and the current fluctuates. The controlled parameter is usually defined by four significant figures, but the measured one typically varies instantaneously over a range of several percent.
When current is passed through an electrolytic system, the heat power output can be considered as the product of the cell current, I, times the sum of a series of voltages equivalent to the recombination enthalpy (e.g. for D2 + 1/2 O2 to D2O) and the Joule heating terms. The latter include the recombination entropy and the current-dependent overpotentials at the two electrodes (the voltages necessary to drive the electrode kinetics at the rate defined by the current), plus the IR drop in the electrolyte (external IR effects ignored). In a gas evolving system, a further term related to the heat of evaporation may be relevant.
With respect to considerations of experimental error in determining whether there is an additional component in the heat balance equation due to an unknown source of heat (e.g. cold fusion), two comments on these voltage equivalents may be made.
First, the enthalpy term depends on whether the cell is open (gases vented, no recombination enthalpy) or closed (total recombination and fully recovered enthalpy). Since most experiments quoting positive effects in steady state measurements have cited a voltage equivalent to the excess heat power less than, or only slightly greater than, the figure of 1.527 V corresponding to full recombination, the degree to which the cell is open or closed is critical . For example, in the original Fleischmann--Pons paper [FLE], only two of nine cells showed more than 1.527 x I in excess heat power, and then only slightly.
Second, the cells listed for both positive heat effects in Table 2.1 typically have voltages in the range of 5-20 V. The larger contributions to this magnitude are the overpotentials and the IR drop. Since these Joule components cause both heating (temperature increment) and power input (voltage increment), they obscure the observation of any experimental difference assignable to an excess heat with a voltage equivalent smaller than or on the order of 1.527 V. Thus their existence effectively multiplies the expected error in the output heat excess determination by factors of ca. (cell voltage - heat excess voltage)/(heat excess voltage), where the denominator is the excess heat expressed as its voltage equivalent.
This factor can obviously be very large and places an increasing burden on the quality of the measurements as the current density rises (implying larger overpotentials and I2R heat). Even an overall precision of 1% of input power at 20 V implies a corresponding relative error for excess heat (even of a magnitude equivalent to 1.5 V) of greater than 10%. With the high rate of gas evolution of cells operating at 500 mA/cm2, instantaneous variation of cell voltages due to both changing overpotentials (actual unblocked surface areas of electrodes vary) and IR drops (fluctuating electrolytic resistive path) can be several percent or more. Such analysis only serves to point out the rather severe requirements of this type of calorimetry attempted in the high current density regimes which have been applied in typical cold fusion experiments.
The calibration method used by most groups reporting positive results is based on the temporary addition of a power increment delta-P via a resistive element in the cell. This delta-P causes a change in the steady-state temperature by an amount T, then a differential heat transfer coefficient kD = delta-P/delta-T is calculated. This measurement can be carried out as the cell is running. The resistive power increment is added on top of that from the electrolytic process, thus maintaining similar stirring conditions for power dissipation. Once kD is determined, the heat power evolved from the electrochemical action is calculated as the product of kD and the difference in temperature between the operating cell (with the calibration heater off) and the external sink. This temperature differential sometimes reaches 30° C or more, but this could develop either from excess heat or from high input current densities leading to large resistive power dissipations in the solution.
In principle and in practice, the differential coefficient kD depends on the magnitude of the difference in temperature between bath and cell. In using the value of kD at the operating point to calculate the evolved heat power, investigators often assume that kD remains constant over the range of cell temperatures from the value of the bath to the operating point. The real need is for the integral heat transfer coefficient kI at the operating temperature Tc, which is truly related to the evolved heat power P as kI(Tc-Tb), where Tb is the bath temperature. The differential value kD, is related to kI, as:
Some groups use small added heats, relative to the electrolytic power, and a single point calibration. Others use values 20-100 times as large as the reported excess heat. In these latter cases, the calibration is approximately a measurement of the integral heat transfer coefficient for a wide span of power above the operating point. This value is then assumed to be equal to the integral heat transfer coefficient below the operating point.
In all its variants, this calibration method is intrinsically prone to overestimate the evolved heat power, because the temperature differential is typically a nonlinear function of heat power in such a manner as to produce a positive error in the estimated kI, (taken as measured kD). An error of a few percent in the assumption that kI=kD would invalidate nearly all reports of excess heat.
The methodology is, in principle, capable of defining the evolved heat accurately. Calibration needs to be done frequently enough over the operating range to be able to calculate kI directly; however the panel saw no instance in which calibrations were made in sufficient detail to allow the calculation of a true kI. If there are changes in the structure of this system (e.g., the electrolyte level in the cell), the heat transfer coefficient becomes time dependent. Repeated recalibration is then required.
A surprising aspect of the calorimetry related to cold fusion is the lack of attention that has been given so generally in reports of excess heat, to the statistical assessment of errors. It is evident on the face of the data in some reports that a group's claim of excess heat is not supported with results of sufficient precision to allow such a conclusion. More usually, it is not possible to assess precision from reported results, because the result is reported from a single run and no error bars are provided for the measured parameters. Conclusions in this arena simply cannot be accepted without a thorough assessment of the measurement errors. In its visits and conversations, the members of the Panel were struck repeatedly by the absence of critical assessments of this kind. Several different kinds of problems were common:
Moreover, errors in calculated results are typically not developed systematically from the measurables via arguments based on propagation of error. Sometimes linear relationships yielding a critical slope or intercept are used in calibration or in the final evaluation of excess heat effects, yet the lower limit of standard deviation of the slope or intercept is not assessed even from the evident scatter in the data.
In some of the laboratories examined, the electrolytic cells were operated at constant voltage, not constant current. Since the conductivity of solutions of LiOD is only about half that of corresponding solutions of LiOH, and since the cells were operated in a regime where the applied voltage appeared as a drop across the solution resistance, cells operated with H2O electrolyzed more moles per second than equivalent cells containing D2O. In some representations of data (e.g. temperature differential vs. applied power for cells in the two different solvents), this effect can show apparent excess heat for the cells evolving deuterium, simply because less power is carried out of the cell in the enthalpy of vaporization of water carried by the vented products.
A careful consideration of measurements with thermocouples clearly illustrates the difficulty of precise measurements of evolved heat power. The excess heats reported by several groups normally correspond to about one half degree in temperature differential. The sensitivity of the thermocouples is typically about 40 microvolts per degree, hence the effect is represented by a dc signal on the order of 20 microvolts, readable with a meter precision of about 2 microvolts per reading, or about 3 microvolts for the difference of two readings. The scatter of data points in various studies suggests that the actual precision is no better than about half to one-third of the differential ascribed to excess heat. Precise measurement of the differential is, of course, only half the story. One also needs accurate measurements of dc voltages at these low levels. If there were an experimental bias inducing even a few microvolts at the thermocouple in the cell, one could see an apparent excess heat. Since the thermocouple is operated in the electric field caused by the resistance losses in solution, such an induced signal is a very real possibility. This problem could also manifest itself at thermistors. One additional source of error is the improper choice of cold junctions in thermocouple measurements.
A thermocouple is normally encased in a glass or ceramic tube containing a heat transfer medium (sometimes water). It is assumed to be electrically isolated from the solution, but a few microvolts of pickup are not beyond the realm of possibility. Since the electric fields are typically larger in D2O solutions than in H2O solutions, this effect could be relevant to the differences between cells operated in D2O vs. H2O. It would also explain why the excess heat power sometimes seems to be nearly linear with applied electrolytic power.
Corrosion of the thermocouple in the heat transfer fluid could also produce dc offsets large enough to yield an apparent excess heat, especially if the heat transfer fluid were a polar liquid such as water.
The precise, accurate measurement of temperature differentials via thermocouples at the level of quality required to confirm typically reported excess heats is an experimental challenge of a high order.
A completely different problem is with the assumption that the input electrical power is simply the product of a time-averaged current and a time-averaged voltage. If an alternating component is present, there can be an error arising from the fact that the time-averaged power is generally not the same as this product. In effect, the measurements, as usually practiced, are blind to the ac component of power dissipation, so that the electrical input power is underestimated and an "excess heat" can be expected in the calorimetry. The fluctuations in mass transfer and surface coverage discussed above produce some ac component in nearly all cells. In addition, there are reports that the current or cell voltage can undergo oscillatory behavior with large amplitudes for brief periods [MOL] upon changing the operating conditions, such as the current density. Such effects would lead to sizable transitory discrepancies between the apparent (dc) and true (ac + dc) electrical input power, hence they might be the basis of some of the bursts that are reported by various groups.
An assumption in most cases for which positive heat effects are reported is that the gaseous products D2 and O2, are vented without recombination. It is widely recognized that any recombination of these products to regenerate deuterium oxide within the cell will produce heat that will appear as an excess, simply because of the correction for the enthalpy of the products. Many groups have guarded against this error by measuring the volumes of gases produced and comparing them with the expectations based on the charge passed through the cell. On the other hand, these measurements are sometimes not made at all, and in other cases they are not made with sufficient precision and accuracy to assure the absence of recombination at a level that could account for the thermal excess. In a surprising number of instances, the cells were operated with Pt or Pd surfaces exposed in the headspace above the solution. These are efficient catalytic surfaces and could be expected to produce a breakdown of the assumption of no recombination.
A related, but different, issue concerns the current efficiencies at the anode and cathode. In open cells, the application of the enthalpic correction implies that 100% of the electron flow at the cathode is devoted to the production of D2 and that the same perfect efficiency applies to the generation of O2 at the anode. If D2 reaches to the anode, it will be oxidized there in competition with the D2O and will invalidate this assumption. The same reduction in efficiency will be seen at the cathode if O2 reaches its surface. The net effect is a kind of "electrochemical recombination" within the cell, and to the extent that it occurs, one cannot validly correct the input power for the enthalpy of vented products in the usual manner. In virtually all cells, the anode and cathode operate without a diaphragm separating them, so some reduction in current efficiency is inevitable. The effect was actually evaluated quantitatively [SHE, CUN]. It is not of concern in a closed cell.