Molex Process: Zeolite Adsorptive Separation

• The Molex process is an adsorptive separation technique that exploits the steric and entropic selectivity of zeolitic micropores to isolate linear hydrocarbons from branched isomers.
• It operates cyclically through adsorption and regeneration under controlled temperature and pressure conditions to achieve ultrahigh product purity.
• Benchmark comparisons indicate that while Molex outperforms fractional distillation for closely related isomers, it incurs higher energy penalties than more advanced separation methods like Levi–Blow.

The Molex process is an industrial adsorptive separation process developed by UOP for separating linear paraffins from branched or cyclic hydrocarbons by shape-selective adsorption in zeolites; Molex is a trademark and/or service mark of UOP Inc. In the hydrocarbon-isomer benchmark of "Energy Efficiency of Hydrocarbon Isomer Separation via Levi-Blow Mechanism: Benchmarking Against Conventional Methods" (Nag et al., 1 Sep 2025), Molex is treated as the conventional adsorptive benchmark for an equimolar neopentane/$n$-pentane feed. In that setting, its defining principle is not vapor–liquid equilibrium, but steric and entropic selectivity in crystalline micropores: linear $n$-pentane can enter and be adsorbed within zeolitic pores, whereas branched neopentane is sterically excluded or much less strongly retained.

1. Industrial role and separation target

In the formulation summarized in (Nag et al., 1 Sep 2025), the Molex process addresses a class of refinery separations in which hydrocarbon isomers have very similar bulk thermophysical properties, making distillation difficult. The neopentane/$n$-pentane pair is presented as a representative case: the two isomers are closely matched in conventional separation-relevant properties, yet they differ in molecular shape and kinetic diameter. Molex exploits that asymmetry by retaining the linear isomer in the adsorbent while allowing the branched isomer to leave in the raffinate.

The process is therefore positioned as a conventional alternative to fractional distillation when boiling-point-based discrimination is weak. In the specific benchmark, the intended product is high-purity neopentane, while $n$-pentane is the retained or adsorbed species. This means that the Molex column functions as a shape-selective molecular sieve rather than as a phase-equilibrium device.

A central implication of this framing is that Molex is most naturally understood as an adsorption-regeneration process for structurally similar hydrocarbons. The paper characterizes it as much better suited than distillation for closely related isomers, but still subject to substantial thermal and regeneration penalties at high purity (Nag et al., 1 Sep 2025).

2. Separation mechanism, adsorbents, and process sequencing

The physical basis of Molex in (Nag et al., 1 Sep 2025) is diffusion and adsorption of linear paraffins into zeolitic micropores, with branched species rejected by steric constraints. The adsorbent class is given as fixed-bed crystalline zeolites such as 5A or NaY, and the paper cites a pore size around $\sim 5.1~\text{\AA}$ for shape-selective uptake of linear alkanes. The selectivity basis is kinetic-diameter and pore-size matching, described explicitly as shape-selective molecular sieving.

Operationally, feed enters the column, linear molecules are retained, and branched molecules leave in the raffinate. Regeneration is carried out by pressure swing or thermal swing desorption, with purge gas and/or temperature ramp. The process is therefore cyclic by construction: adsorption enriches the desired branched product stream, while desorption restores adsorbent capacity for the next pass (Nag et al., 1 Sep 2025).

In the neopentane/$n$-pentane case, the paper emphasizes that linear $n$-pentane can enter and be adsorbed within the pores, whereas branched neopentane is sterically excluded or much less strongly retained. This selective accessibility of microporous adsorption sites is the core of the Molex mechanism. A plausible implication is that process performance depends not only on equilibrium selectivity but also on steric accessibility and intracrystalline transport; however, the paper does not provide adsorption isotherms, breakthrough curves, or explicit mass-transfer coefficients.

3. Benchmark operating conditions and purity progression

The benchmarking study does not report new Molex experiments. Instead, it uses a literature-informed, first-principles estimate for an initially equimolar feed,

$x_{\text{neo}}=0.5,\qquad x_{n\text{-pent}}=0.5,$

with operation assumed at about $170^\circ\text{C}$ ($443~\text{K}$) and $n$0–$n$1, after heating the feed from $n$2 (Nag et al., 1 Sep 2025).

One cycle is defined as one adsorption-based separation pass. The paper assumes that a single pass produces a neopentane product stream of approximately $n$3 neopentane and $n$4 $n$5-pentane. That one-pass selectivity is then iterated to estimate how many repeated cycles are needed to reach increasingly stringent purity targets, labeled as one-nine, three-nine, four-nine, six-nine, and eight-nine purity.

The purity bookkeeping uses $n$6 and $n$7 for the moles of neopentane and $n$8-pentane in the product pool after each iteration, with normalized mole fractions

$n$9

Within this model, repeated $n$0 cleanup yields ultrahigh purity in a small number of cycles.

Iteration	$n$1	$n$2 (kJ/mol)
0	0.5	0.0
1	0.98	51.87
2 (one-9)	0.999	103.74
3 (three-9)	0.99999	155.61
4 (four-9)	0.999999	207.48
5 (six-9)	0.99999999	259.35
6 (eight-9)	0.9999999999	311.22

According to this iterative benchmark, Molex reaches eight-9 purity in 6 cycles. The paper also states that after 6 cycles the retained neopentane amount is $n$3, corresponding to $n$4 recovery relative to the initial $n$5 neopentane in the feed. On that basis, the cumulative $n$6 is normalized to

$n$7

and the cycle count is similarly normalized to about $n$8 cycles per mole of purified product (Nag et al., 1 Sep 2025).

The paper also notes that the purity labels and decimal values are not always perfectly aligned. This suggests that the principal result is the iterative trend toward ultrahigh purity, rather than the exact nomenclature assigned to each purity level.

4. Energy model and thermodynamic interpretation

The Molex benchmark is constructed from two layers of energy accounting. The first is a cycle-by-cycle hydrocarbon-heating estimate. The second is a much larger whole-column thermal estimate that includes heating the entire zeolite bed (Nag et al., 1 Sep 2025).

For neopentane, treated as already gaseous at $n$9, the paper gives

$n$0

with the numerical value

$n$1

For $n$2-pentane, the paper includes liquid heating, vaporization, and gas-phase heating: $n$3 with numerical evaluation

$n$4

The text then states

$n$5

This is numerically inconsistent with the preceding $n$6 and $n$7 values. The cycle table introduces a second inconsistency by using $n$8 as the cumulative increment per cycle. The paper nonetheless uses the tabulated values in the final normalized Molex comparison. The coexistence of $n$9 in the narrative and $\sim 5.1~\text{\AA}$0 in the iterative table is one of the study’s explicit accounting ambiguities.

To place these values against a reversible limit, the paper evaluates the thermodynamic minimum work of separation for an ideal equimolar binary mixture: $\sim 5.1~\text{\AA}$1 For $\sim 5.1~\text{\AA}$2 at $\sim 5.1~\text{\AA}$3, it gives

$\sim 5.1~\text{\AA}$4

Using the normalized Molex value $\sim 5.1~\text{\AA}$5, the paper reports an efficiency of $\sim 5.1~\text{\AA}$6, evidently computed as

$\sim 5.1~\text{\AA}$7

and equivalently places Molex at roughly

$\sim 5.1~\text{\AA}$8

times the thermodynamic minimum. This is the origin of the characterization of Molex as approximately $\sim 5.1~\text{\AA}$9 above the lower bound (Nag et al., 1 Sep 2025).

5. Comparison with fractional distillation and Levi–Blow

For the same neopentane/$n$0-pentane separation at highest purity, the paper compares fractional distillation, Molex, and the Levi–Blow method. Using the normalized-product accounting, it reports the following values: fractional distillation at $n$1 and $n$2 cycles, Molex at $n$3 and about $n$4 normalized cycles, and Levi–Blow at $n$5 and $n$6 cycle (Nag et al., 1 Sep 2025).

This comparison underpins the paper’s broad conclusion that Molex is vastly better than fractional distillation for this isomer pair, but still an order of magnitude worse than Levi–Blow in energy. In abstract form, the study states that Levi–Blow operates within a factor of $n$7 of the thermodynamic minimum, Molex at $n$8, and fractional distillation at $n$9.

A second, more severe Molex energy figure appears when the paper incorporates the thermal load associated with heating the full zeolite bed. For a cylindrical zeolite column of approximately $n$0, with zeolite specific heat $n$1, the paper estimates

$n$2

and obtains

$n$3

This whole-column heating estimate is the origin of the paper’s statement that Molex can demand on the order of $n$4 at high purity, because bulk thermal load dominates the energy budget.

The study does not fully reconcile these two Molex views. One is the cycle-by-cycle hydrocarbon-processing benchmark, leading to $n$5. The other includes bulk zeolite-bed heating and yields an energy burden on the order of $n$6. The abstract’s $n$7 factor relative to minimum work is consistent only with the lower Molex figure, not with the whole-column heating estimate.

6. Limitations, interpretation, and nomenclature

The paper’s treatment of Molex is explicitly a benchmarking estimate rather than a rigorous industrial process design. It does not present new Molex experiments, and it does not provide adsorption isotherms, breakthrough calculations, detailed PSA or TSA scheduling, bed mass-transfer models, or industrial process optimization (Nag et al., 1 Sep 2025). The quantitative results therefore function as a comparative first-principles benchmark rather than as a definitive simulation of refinery operation.

Several caveats are stated directly. The energy accounting is internally inconsistent in places; the purity labels do not always align exactly with the reported decimal purities; the feed is idealized as only neopentane and $n$8-pentane; the one-pass Molex separation is idealized as $n$9; and the bed-heating estimate is coarse. These caveats do not negate the paper’s qualitative ranking, but they constrain the precision with which its Molex figures should be interpreted.

In the paper’s synthesized view, the principal advantages of Molex are that it is industrially established, that it leverages shape-selective adsorption rather than weak boiling-point differences, and that it reaches eight-9 purity in $x_{\text{neo}}=0.5,\qquad x_{n\text{-pent}}=0.5,$0 cycles rather than the more than $x_{\text{neo}}=0.5,\qquad x_{n\text{-pent}}=0.5,$1 iterations quoted for fractional distillation. Its limitations are the need for elevated temperature, moderate pressure, repeated adsorption–desorption cycling, and a potentially dominant bulk thermal penalty when the entire zeolite bed must be heated.

A separate nomenclature issue arises in contemporary arXiv usage. Near-homographic names denote unrelated machine-learning methods, including "MoLEx: Mixture of Layer Experts for Finetuning with Sparse Upcycling" (Teo et al., 14 Mar 2025) and "Unveiling Molecular Secrets: An LLM-Augmented Linear Model for Explainable and Calibratable Molecular Property Prediction" (Li et al., 2024). These are not the petrochemical Molex process. This suggests that capitalization and domain context are important when searching the literature.

Overall, the Molex process, as represented in (Nag et al., 1 Sep 2025), is a zeolite-based adsorptive benchmark for separating linear hydrocarbons from branched isomers by steric and entropic selectivity in micropores. For the modeled equimolar neopentane/$x_{\text{neo}}=0.5,\qquad x_{n\text{-pent}}=0.5,$2-pentane system, it is reported to achieve ultrahigh neopentane purity after repeated cycles and to perform dramatically better than fractional distillation, while remaining substantially less energy-efficient than the proposed Levi–Blow method.