[Editor's Note: The design study underlying this article is the author's master's thesis, "Technical and Economic Proposal to Remove the H₂S Content From the Residual Gas From Process Plants Operating in a Refinery in Colombia," completed at Universidad Viña del Mar. The facility is presented anonymously.]
Three years from now, someone will ask why you chose the technology you chose. It might be a regulator during a permit renewal, an auditor tracing an emissions figure back to its basis, or an engineer who has just inherited your unit. The question is rarely whether the decision was correct. It is whether you can show how it was reached. Most technology-selection files cannot, because the justification was written after the choice rather than assembled during it.
This article describes a five-criterion framework that approaches sulfur-removal selection in the reverse of the conventional order. It begins with the governing emission limit and builds forward, creating a documented decision record before a technology is selected. The worked example is a fuel-gas-sweetening retrofit at a refinery. The useful part is not which technology won, but what the process revealed along the way, including a constraint that had nothing to do with chemistry.
Why the Order Matters
The conventional sequence is familiar to anyone who has done it: shortlist two or three technologies on removal efficiency, compare installed cost, choose, then write the justification memo. That works, in the sense that it produces a decision.
What it does not produce is a record. By the time the memo is written, the inputs have been rounded, the discarded options are described from memory, and the reason a particular assumption was acceptable has stopped being written down. The cost of that appears later, when the plant changes and someone has to decide whether the original basis still holds. If you cannot tell which numbers were measured and which were estimated, you cannot answer that question. You can only redo the study.
Five Criteria, In Order
Reversing the sequence means the limit comes first, and every candidate is then evaluated against the same five criteria in the same order (Fig. 1).
Compliance margin comes first, ahead of cost. The question is not only whether a technology can meet the limit, but how much operating margin it provides and and whether that margin is stable across the operating cycle.
Equipment count and process complexity serve as a proxy for everything that follows from having more equipment: more failure modes, greater maintenance exposure, more people who need to understand the unit, and higher capital costs.
Safety and process-risk exposure covers the hazards each option introduces rather than the one it removes—combustion systems, alkaline-solution handling, and spill and containment pathways.
Sulfur removed vs. sulfur recovered separates two things that are easy to conflate. Recovery produces a saleable product and a revenue line. Removal produces a spent-media disposal stream. Which is appropriate depends almost entirely on scale, and naming the distinction as its own criterion forces that conversation early.
Life cycle cost comes last by design. Put it first and it dominates the decision; put it fifth and it decides among options that have already survived on their merits.
The Retrofit
The case was a refinery processing 250,000 B/D of crude. A residual gas stream used as plant fuel carried 1,067.5 ppm H2S at 1,383,370 SCFD, with unusually low CO2 at 0.011 mol%. Corrosion was visible at burner nozzles and in boiler sections. The governing constraint was the national industrial SO2 emission standard, which capped emissions at 550 mg/m3—roughly 20 ppm.
Four options went through all five criteria: the Claus process, the Stretford process, CrystaSulf (a patented nonregenerative liquid redox process that removes H2S from gas streams and converts it directly into elemental sulfur), and a nonregenerative iron-oxide dry sorbent bed. Dry-bed sweetening is conventionally an upstream, wellhead-adjacent technology, and its behavior in fixed-bed service is well documented—see SPE 13280 and PetroWiki's overview of sour gas sweetening. Applying it downstream, in a refinery fuel-gas network, was the part that needed justifying.
What the Criteria Surfaced That Cost Alone Did Not
Criterion 1 did the first real work. A liquid scavenger meets the limit on day one, but its outlet concentration climbs as the reagent is consumed, so the compliance margin decays through the changeout cycle. The dry bed holds its outlet essentially flat across bed life. Both technologies pass; only one passes the same way in week 11 as in week one. The trade-offs between the two approaches are laid out in detail in NAICE 16815 and SPE 213604.
Criterion 4 eliminated Claus, and it did so on scale rather than on price. The stream required removal of about 82 kg of H2S per day. A Claus train is a sulfur-production facility; below a certain sulfur load there is no product volume to justify the plant that makes it. Had cost been the first criterion, Claus would have been discarded anyway—but for the wrong reason, and the file would have recorded “too expensive” instead of “wrong process for this duty.” Those are different findings, and only one of them stays true if prices move.
Criterion 2 then separated the survivors. Claus requires a catalytic converter, a condenser, and a combustion furnace. Stretford requires circulation and alkaline solution handling. The dry bed requires a vessel, a bed, and instrumentation.
Only after all of that did cost enter (Fig. 2). The dry bed came in 83% below Claus and 71% below Stretford on installed capital. What the cost structure shows is more interesting than the totals: the dry bed carries the highest media cost of the four options and still lands lowest overall. The saving is in equipment count, which is what criterion 2 had already predicted.
The Constraint That Actually Decided the Design
With the technology settled, sizing looked like a straightforward optimization: Longer media life means fewer changeouts, so the longest-lasting sorbent that meets the specification should be the obvious choice. In practice, it is not that simple, and the reason is geometric.
More capacity means more media, and in a fixed-bed column more media means a taller bed. Fig. 3 plots required bed height against changeout interval for two candidate diameters. Moving from 90-day to 180-day media at a 1.6-m diameter takes the bed from 14 m to more than 50 m. Every point on both curves meets the outlet specification. The chemistry does not choose between them. Available headroom does.
The slenderness ratio (height/diameter, H/D) is the practical design guide, with an efficient band around 5 to 10. Below this range, the vessel becomes uneconomically wide and the bed too shallow for the gas to distribute evenly across it; above it, pressure drop rises in proportion to bed height, increasing compression duty and narrowing the margin to the fluidization limit. The selected configuration was 90-day media in a 1.6-m column with a-14.05 m bed, giving H/D of 8.8. A minimum-fluidization check confirmed the bed stays fixed in upflow operation because the calculated minimum fluidization velocity (Umf) of 0.1625 m/s is well above the operating superficial velocity. For readers interested in the underlying transport treatment, the shrinking-core model for fixed-bed H2S adsorption towers is a good starting point.
The lesson generalizes beyond sweetening. The binding constraint on a retrofit is very often geometric or logistical rather than chemical, and it is worth looking for it before optimizing anything else.
Label Provenance While You Still Know It
The practice that makes the rest of this work durable is unglamorous: every input in the study is labeled according to its source, whether measured, vendor-supplied, estimated, or assumed (Fig. 4).
It costs a single column in a spreadsheet;. the payoff arrives years later. Now 9 years old, the illustrative study and the labels make its condition legible at a glance: the measured gas composition and flow are still valid, the vendor loading ratios can be reconfirmed from a current data sheet, and the cost and currency lines are visibly outdated and need replacing before anyone relies on them. Without the labels, all of those numbers would look equally authoritative, and requalifying the study would mean starting over.
That is the difference between a decision that was merely correct and one that can be re-derived. The latter is worth considerably more.
Lessons Worth Carrying to the Next Selection
- Start at the limit, not the shortlist. The requirement is the only fixed point in the problem. Everything else is a candidate.
- Put cost last. Not because it does not matter, but because it will win every argument if introduced too early.
- Distinguish removal from recovery. Scale decides which is appropriate, and the reasoning survives price changes in a way that a cost comparison does not.
- Look for the binding constraint before optimizing. It is frequently not the one your discipline trains you to look at.
- Label provenance as you go. You will never again know as clearly as you do today which numbers you measured and which you assumed.
None of this requires software or a formal management system. It requires deciding, before the analysis starts, what the record will need to contain—and then keeping it as the work proceeds rather than reconstructing it afterward.
For Further Reading
SPE 13280 An Updated Examination of Gas Sweetening by the Iron Sponge Process by J. Anerousis and K. Whitman, Physichem Technologies Inc.
NAICE 16815 Liquid Scavenger vs. Fixed Bed H2S Adsorbent. Working in Harmony or Against Each Other for H2S Removal by S. Lim, A. Jenkins, K. Barbuto, SLB.
SPE 213604 A Hybrid H2S Removal Solution. Using Liquid vs. Fixed Bed H2S Scavengers in Harmony by M. Crawshaw, W. Brundick, M. Juncker, et al., SLB.
Shrinking-Core Model Integrating to the Fluid-Dynamic Analysis of Fixed-Bed Adsorption Towers for H2S Removal from Natural Gas by B. Carrasco, E. Ávila, A. Viloria, et al., Yachay Tech University.
