Karl Fischer titration: titre, drift and what a water result really means

A Karl Fischer result that comes out high is usually not an instrument fault. The three things that produce it, in rough order of frequency, are a titre that has moved since it was last determined, a drift the method is not correcting for, and a sample that picked up atmospheric moisture between the weighing room and the titration cell. All three are operator-controlled, and none of them produces an error message.

The chemistry is the same across the standards: water reacts quantitatively with iodine and sulfur dioxide in an anhydrous medium containing a base with sufficient buffering capacity. ISO 760 describes the general method, with visual or electrometric end-point detection and the electrometric version recommended as the more accurate. USP <921> gives Method I in a direct form (Ia) and a residual form (Ib). Ph. Eur. splits the two techniques: 2.5.12 for the semi-micro volumetric determination and 2.5.32 for the micro, coulometric determination.

Volumetric or coulometric: a method choice, not a preference

ASTM E203, which covers the volumetric method, makes the boundary explicit: it notes that where samples routinely contain water at about 500 mg/kg or less the coulometric technique should be considered, because it is more accurate at those levels. Running a trace-level sample volumetrically means titrating a few microlitres of titrant and living with the burette's resolution as a dominant uncertainty.

The titre is the calibration

In volumetric work the whole result rests on one number: the mass of water consumed by one millilitre of titrant. It is not a property of the bottle. Titrant loses strength through exposure to atmospheric moisture, through evaporation of solvent from the reagent bottle, and simply with time, so the titre determined on Monday is not the titre on Friday.

Three standards are used to determine it. Pure water, delivered by syringe or by weight, is the most direct and the most demanding, because delivering ten microlitres accurately is harder than it sounds. A certified water standard — a solution of stated water content with an uncertainty on its certificate — removes the delivery problem and is what most laboratories use. A crystalline hydrate, usually sodium tartrate dihydrate, is the convenient solid option: its water content follows from the formula, two molecules of water in a formula mass of about 230, which works out at roughly 15.66 % by mass. That figure is arithmetic from the stoichiometry rather than an assigned value, and it holds only if the material really is the dihydrate and has not partly dehydrated in a warm cupboard.

Replicate the titre determination and look at the scatter. A titre from a single determination has no repeatability attached to it, and since it multiplies every subsequent result, its uncertainty propagates into every water content the instrument reports until the next determination.

Drift, and the blank nobody subtracts

A sealed titration cell is not sealed. Water enters through the septum, past the ground joints, through the desiccant if it is exhausted, and on the outside of every syringe needle that goes in. The rate at which it enters is the drift, expressed as micrograms of water per minute, and a well-maintained cell on a dry day sits far below one on a humid afternoon with a tired molecular sieve.

The consequence is systematic and in one direction. A titration that takes four minutes accumulates four minutes of drift as apparent sample water. For a 50 mg sample at 0.1 % water — 50 micrograms of water — a drift of 10 micrograms per minute over four minutes is forty micrograms, which is most of the result. Drift correction exists for precisely this reason, and it is the first setting to check when a method gives high results on small samples.

• Let the drift stabilise before titrating. A cell that has just been filled is still drying itself.
• Record the drift value with the result. It is diagnostic data: a rising trend across a week means a seal, a septum or a desiccant, not a sample.
• Replace the septum on a schedule rather than when it visibly fails — a needle-punctured septum leaks long before it looks wrong.
• Keep solvent volume within the working range of the cell. An exhausted solvent slows the reaction and lengthens the titration, which multiplies the drift.
• Run a blank by the same procedure with no sample, so the method's own background is measured rather than assumed to be zero.

Interference and side reactions

Karl Fischer is selective for water but not immune to chemistry. Ketones and aldehydes react with methanol in the reagent to form acetals or ketals, liberating water and giving a result that climbs without ever reaching a stable end point — the classic drifting, never-finishing titration. Reagent formulations intended for aldehydes and ketones exist and use a different solvent system; the fix is the right reagent, not a longer wait.

Other cases: strongly basic samples push the pH outside the buffered working range and slow the reaction; strongly acidic ones do the same in the other direction; carbonates and bicarbonates liberate water on reaction with the acidic medium; some oxidising and reducing species consume iodine or iodide directly. Where the chemistry is in doubt, the honest test is a recovery experiment — spike a known quantity of water into the matrix and see whether you get it back.

Sample handling is most of the measurement

For a sample at 0.1 % water, a 50 mg portion contains 50 micrograms of water. A fingerprint, a few seconds in humid room air, or a syringe rinsed with wet methanol can all contribute at that scale. This is why hygroscopic materials are weighed into sealed vials and introduced by septum, why the needle is wiped, and why some laboratories weigh by difference with the vial closed between weighings.

It is also why a water result and a balance calibration are connected. The water content is a mass fraction, so the sample mass divides into the answer directly, and a sample weighed below the balance's minimum weight carries a relative uncertainty that no amount of titration care recovers. For small samples the balance is often the larger contributor to the budget.

Oven and headspace methods

Where a sample cannot go into the cell — because it dissolves badly, reacts with the reagent, or releases its water only on heating — the sample is heated in a separate vessel and the evolved moisture is carried into the titration cell by a dry gas stream. This removes the matrix from the chemistry entirely, which is its attraction, and introduces two new variables: the oven temperature has to be high enough to drive water out and low enough not to decompose the sample into something that also releases water, and the transfer line has to be dry and leak-free along its whole length.

A temperature study is therefore part of developing an oven method. Water evolved against temperature should reach a plateau; a result that keeps climbing as the temperature rises is decomposition being reported as moisture.

ValiTrac AI is not an accredited laboratory and confers no compliance. Water content limits are set by the monograph or specification for the material, and the acceptance criteria for titre determination, drift and recovery belong to your own method and quality system.