Caustic Layering

the forgotten hazard

Ineffective dissolution of solid caustic soda presents a risk of eruption, reports John Cox

THE PHENOMENON of "Caustic Layering", first identified in the 19th. Century, is rarely encountered today. Suppliers of caustic soda have produced excellent technical literature with clear warnings of the danger of creating weak and strong layers during dissolution:

"It is important not to add the pearls too quickly or they will tend to fall to the bottom and form a large agglomerate that will be slow to dissolve. Agitation is also important to prevent the formation of a concentrated layer of solution which would suddenly mix with a layer of weaker solution and cause violent boiling".1

The boiling risk may be evaluated using the enthalpy-composition diagram2 commonly used to calculate the heat evolved when diluting a caustic soda solution. This is illustrated on Fig.23 by the dotted line joining B1 (water at 65C) to B2 (a 55% solution at 77C). The line traces all the intermediate mixtures (Point A, for example, with a composition of 27% NaOH, corresponds to a mixture of equal volumes of water and the 55% NaOH solution).

This procedure not only enables the temperature of a mixture to be determined (for example, 112C for Point A) but also, if a 'Mixing line' cuts the 'Boiling Point curve', to indicate whether boiling will occur during mixing. This is important because, although a simple 'splashback' is an ever-present hazard, a major eruption is only possible with boiling.

The enthalpy-composition relationship also controls the behaviour of strong and weak layers during mixing. In this instance however, because both layers are already in the vessel, the mixing is more rapid. This is why, for an equivalent energy release, the force of an eruption from the mixing of layers is greater than from dilution or dissolution (which, normally, takes place over a period of many seconds).

Although Suppliers are fully aware of this hazard, it remains a mystery to some Customers. As an illustration, a specific incident at a Resins Plant, and the subsequent investigation and High Court litigation, is described in detail. The Company's RIDDOR report (8th. July 1987) made their position clear from Day 1.

"An operator tending a resin reactor (being caustic cleaned) reported to his shift manager having been splashed with caustic soda solution. After preliminary treatment, drenching and irrigation, he has been taken to hospital where concern has been expressed over his eyes. No witness saw the accident but he has probably charged a bag of caustic soda into the reactor containing hot caustic solution and been splashed by the reactor".

This initial finding was endorsed by a formal investigation which relied on an interpretation of the temperature record (Chart 1). This showed a 3C rise at 06.30am coinciding with whatever caused the eruption and subsequent injuries to the process worker. As an extra 25kgs. of caustic soda would have caused a 2.7C rise, it was concluded that a 25kg. bag must have been added.

This was already the consensus when the process worker was interviewed in hospital some weeks later. No-one believed his claim to have opened the inspection cover for a "quick look inside" and that, after a "split-second", the vessel erupted spontaneously. When he sought compensation, his Solicitors were informed:

"The accident occurred because contrary to (his) supervisors instructions, he added solid caustic soda to a hot caustic cleaning solution in a reactor. This caused an eruption through the manhole which resulted in injury to your client. My client provided protective goggles and equipment for your client. Your client was the author of his own misfortune and I am therefore not able to advise my client to compensate (him).

Despite this rejection, the process worker persisted with his story. He said that he had had no problem releasing and lifting the strong-arm but some difficulty with a "sticky seal" and that, when he lifted "the cover slightly" (for a quick look inside), "the lid shot back" and fluid splashed with such force as to "penetrate his protective clothing and by-pass his goggles".

The only hard evidence to substantiate this account was his injuries - marked red on the drawings. These were clearly more credible for the posture he described than that required if he had been discharging a bag of caustic soda - notably:

injuries to the left inside arm: none to the right arm

injuries to the right knuckles: none to the left hand

ingress of caustic into the goggles:

more on left side

greater facial injury on the left side and scalp

no injuries to the torso or legs.

Although this should have led the Company to consider whether the eruption could have resulted from the disturbance of a layered solution, they did not do so. This is astonishing since even a cursory scrutiny of the temperature record (Chart 1) reveals abnormal pre-accident operations by the previous night-shift.

The basic procedures were supposed to be:

  firstly, to charge 2500 litres of cool water (18C that day)

  then, to set the temperature controller to not exceed ~37C - and, about the same time, to start the agitator

  then to charge 8 25kg. bags of caustic soda (7 that day)

  then to steam-heat to ~80C (77C that day) (with a routine break at ~70C to test the cooling system)

With hindsight, it is instructive to compare Chart 1 with a simulation (for the subsequent litigation) of these same conditions (Chart 2). Three abnormalities of Chart 1 (all absent from Chart 2) should have alerted the investigation team.

    The initial temperature rise took only 2 minutes - under 25 seconds for each bag (if all seven were emptied in this period), more than twice the normal charging rate and even faster than the 3C jump at 06:30 (c.f. 40 seconds/bag for the simulation).

    The rise stopped at 34C and then meandered to 37C - without any of the features normally associated with a controlled variable in an agitated vessel. Furthermore, the cooling check at 70C did not produce a temperature drop (c.f. Chart 2 - which did).

    If all 175kgs. dissolved in 2500 litres without cooling, the temperature should have reached 37C by 03:20am (q.v. Fig.1).

These pre-accident abnormalities could be explained if:

      Less than 2500 litres water were present as the first bags of caustic soda were charged at an abnormally fast rate.

      The vessel's agitator was not operating at that time.

      About 40kgs. of caustic soda remained undissolved and, later, formed an agglomerate at the bottom of the reactor.

This is a classic scenario for the creation of layered solutions and the investigation team should have realised the implication. They were equally blind to the shortcomings of their preferred theory. Having calculated that a 2.7C rise would result from adding 25kgs. to 2500 litres, they ignored the corollary that the actual rise of 3C could not be explained by the addition of a single 25kg. bag.

Fig. 1 (derived from vendors' data4 and Chart 1) reveals another problem: that boiling would not occur if only one extra bag were charged and that no less than nine bags would have to be added to reach the 16% concentration required for a major eruption.

At the very most, the company's hypothesis could have explained a large spatter of droplets as a recoil from the addition of solid caustic soda. It could not account for a coherent splash or explain why, with the process worker in the posture needed to empty a bag, this would have precisely targeted his head and upper torso or had the force to penetrate the his cotton overalls and gloves or to enter his goggles after hitting his forehead.

In contrast, the events are easily explained if the act of opening the inspection cover disturbed, and triggered the mixing of, a layered solution. An illustrative calculation was submitted in the High Court to show that the process worker's recollections were consistent with this scenario (the 'layering hypothesis').

After the incident, the temperature rose to ~80C and, with 175kgs. of caustic soda dissolved in 2500 litres, its concentration would have been 7% (represented on Fig. 2 by Point D). Before this, there would have been two layers at 77C, at concentrations such that the straight line joining the two points on Fig. 2 would pass through Point D. [The chart reading was actually nearer to 76C but, to avoid an unproductive cross-examination in the High Court, 77C was used to illustrate the argument.]

75% is the maximum concentration of a caustic soda solution at 77C so, if the agglomerate dissolved slowly at this temperature, it could have become a dense 75% layer by 06:30am. For the purposes the High Court presentation, a lower concentration was used (70% - represented by Point C2 on Fig. 2). With this value for the heavier layer, the lighter layer (Point C1, where the extended line from C2 to D cuts the 77C curve) would have had a concentration of 5.4%.

With 5.4% and 70% for the initial concentrations, the volume ratio between the light and dense layers is ~40:1 (obtained by the 'lever rule'). In this instance, the heavier layer would occupy ~60 litres and contain ~42kgs. of the initial 175kgs. charge of caustic soda.

It is noteworthy that 60 litres is less than half the volume of the lower dished end of the reactor and that the interface between the layers would have been only a quarter of the cross-sectional area of the main body of the reactor. In the absence of agitation, this would be an inherently stable situation - until the cover was opened suddenly and a shock wave disturbed the interface.

On Fig. 2, the line between C1 and C2 represents the temperatures of all intermediate concentrations during mixing. This shows that a disturbance at the interface would create transient concentrations in the range from 42% to 18% (where boiling temperatures would be exceeded), that vapour bubbles would form and, within fractions of seconds, the entire contents of the reactor would be mixed.

Although the resultant energy release would have been comparable with that of the Company's hypothesis, the consequences would have been far more serious because the energy release would have been virtually instantaneous and vaporisation would take place at the base of the reactor.

In theory (if vaporisation was truly instantaneous) the volume of the eruption could exceed 5000 litres - twice the volume of the reactor! In practice of course, the mixing and vaporisation would occur over finite period and each bubble would be quickly quenched by its surrounding liquid. The 'in theory' calculation is mentioned simply to indicate the potential volume that could be released by an eruption resulting from caustic layering.

These calculations were presented to the High Court when the case was heard in January 1998 (more than ten years after the incident). There was no cross examination or challenge to this expert opinion since, before the proceedings recommenced, the Company agreed to a 'substantial settlement' - though, as customary on such occasions, without admitting any liability.

John Cox is a Consultant Chemical Engineer who has acted as an expert witness for several process workers injured in accidents at chemical works.



1 "Anhydrous Caustic Soda", ICI Chlor-Chemicals, page 8

2 "Caustic Soda Liquor", ICI Chlor-Chemicals, Fig. 9

3 Derived from Reference 2 but only showing the curve for 77C.

4 Reference 1, page 13