The Gold Standard of Zeolites | Separating Equilibrium Loading from Approach Rate in a Linear-Driving-Force Model

Let q(t) mean amount adsorbed per gram of a specified dry material at time t. We will use mmol/g for loading and minutes for time. A record should also state the adsorbate, material identity, surrounding conditions and initial state. Without those definitions, two numbers carrying the same unit may still describe different objects.

We distinguish three quantities from the start. The loading q(t) is the state at the named time. The equilibrium loading qe is the endpoint approached under the fixed conditions of the model. The rate dq/dt describes how quickly the state is changing. A specimen can have a large qe and a small early loading if its approach is slow.

This distinction extends the Zeolites root’s warning that amount and time answer different questions. Here the task is mathematical: determine which combinations of endpoint and rate can reproduce a specified time record. We will not infer a complete mechanism from the shape of one curve or assume that an equilibrium model alone provides a process timetable.

Contents · Next chapter

For this exercise, assume dq/dt = k(qe − q), with constant qe and positive k. The phrase driving force refers here to the gap between the current average loading and the fixed equilibrium loading. The model says that the instantaneous rate is proportional to that gap.

We stipulate that conditions surrounding the specimen remain fixed in the way needed for qe to stay constant. We also begin with q(0) = 0. The same material basis is used at every time. Measurement response, temperature variation, depletion of a finite reservoir and changes in accessibility are excluded from the initial model.

Those exclusions define the calculation; they are not facts established merely by fitting an exponential. Real adsorption studies may need more detailed transport, thermal or measurement models. A simple equation is useful when its job is clear. The learner should be able to say which assumption would fail if the reservoir changed or if the first reported time occurred after an unrecorded period of uptake.

Contents · Previous chapter · Next chapter

The left side dq/dt has units mmol per gram per minute. The difference qe − q has units mmol/g. Therefore k must have units per minute. It is not a loading, a gas pressure or a diffusion coefficient with units of area per time.

Changing the time unit changes k’s numerical value. A coefficient of 0.6 per minute corresponds to 0.01 per second. The product kt remains the same when time and coefficient are converted consistently. A graph plotted in seconds cannot use the old per-minute coefficient unchanged.

The endpoint qe also retains its reporting basis. If a loading is expressed per gram of finished granules rather than per gram of a specified active fraction, the numerical endpoint may change. That reporting conversion does not by itself identify a different time constant. Keep the time basis and material basis visible in the same record, so a reader can distinguish a unit conversion from a proposed change in the underlying uptake behaviour.

Contents · Previous chapter · Next chapter

Define the remaining gap g(t) = qe − q(t). Since qe is constant, dg/dt = −dq/dt = −kg. A quantity whose rate is a fixed negative multiple of itself follows exponential decay: g(t) = g(0)e^(−kt).

With zero initial loading, g(0) = qe. Hence q(t) = qe[1 − e^(−kt)]. Substitution checks the solution: differentiating gives k qe e^(−kt), while k(qe − q) gives exactly the same expression. At t = 0, the solution returns zero as required.

The formula separates the endpoint scale qe from the fractional approach 1 − e^(−kt). At a fixed positive time, increasing either parameter increases q(t), but for different reasons. That is why one observed loading can leave a trade-off between them. Before trying to estimate two unknown quantities, ask how many independent relationships the observations actually supply and whether those relationships remain distinct under the model.

Contents · Previous chapter · Next chapter

Choose fictional specimen A with qe = 8 mmol/g and k = ln 2 per minute. Then e^(−k) = one half over one minute. At successive whole minutes, the remaining gap halves: eight, four, two, one and one half mmol/g.

The corresponding loadings are zero at the start, four after one minute, six after two minutes, seven after three minutes and 7.5 after four minutes. Each increment is smaller than the previous increment because the unfilled gap is smaller. The coefficient k stayed constant throughout.

Check the record using gaps rather than loadings alone. Eight minus six is two at the second minute, and eight minus seven is one at the third. Their ratio is one half. The loading ratio seven divided by six is not one half, and should not be inserted into the gap-decay equation. Recognising which quantity follows the exponential is an essential part of the method, rather than a cosmetic choice of graph axis.

Contents · Previous chapter · Next chapter

Now choose specimen B with qe = 12 mmol/g and k = ln(3/2) per minute. Its one-minute remaining-gap fraction is two thirds. The loading after one minute is 12(1 − 2/3) = 4 mmol/g, exactly the same as specimen A.

The endpoint is nevertheless different: twelve rather than eight. The approach coefficient is smaller: ln(3/2) rather than ln 2 per minute. At two minutes, specimen B reaches 12(1 − 4/9) = 20/3 mmol/g, approximately 6.667, whereas A reaches six.

This is an exact counterexample to the claim that equal first readings establish equal equilibrium capacity. It does not rely on measurement noise or an inaccurate instrument. Both curves satisfy the same form of equation and the same zero initial condition. The ambiguity comes from observing only one point of a model with two unknown parameters. A better measurement of that same point would not, by itself, reveal which endpoint and rate produced it.

Contents · Previous chapter · Next chapter

Suppose the only supplied observation is q(t1) = Q at a positive known time t1, with Q positive. For any chosen equilibrium loading qe greater than Q, the model can match the point by setting k = −ln(1 − Q/qe)/t1.

The logarithm has an admissible argument between zero and one, so k is positive. Each permitted qe therefore supplies a corresponding rate coefficient. One observation leaves an entire family, rather than merely two rival examples. The earlier specimens are two members of this family for Q = 4 and t1 = one minute.

The boundary deserves attention. Setting qe = Q would require a zero remaining gap at finite time, which the positive finite-k exponential does not reach from zero. Choosing qe below Q is incompatible with monotonic approach from below under this model. These restrictions are useful, but they do not identify one unique endpoint. Distinguish a feasible parameter range from a completed estimate, and preserve that distinction when describing the result to someone choosing the next measurement.

Contents · Previous chapter · Next chapter

Assume that time zero and zero initial loading are known, and observe Q1 = q(Δ) and Q2 = q(2Δ), with Δ positive. Define r = e^(−kΔ). Then Q1 = qe(1 − r) and Q2 = qe(1 − r²).

Factor the second expression as qe(1 − r)(1 + r). Dividing by positive Q1 gives Q2/Q1 = 1 + r. Thus r = Q2/Q1 − 1, k = −ln r/Δ, and qe = Q1/(1 − r).

The endpoint can also be written qe = Q1²/(2Q1 − Q2). These are consequences of the stipulated model and timing, not general formulas for any pair of adsorption measurements. If the second reading is not taken at twice the first time from the known zero start, this particular factorisation does not apply. Write the observation times before choosing an inverse formula; otherwise a correct algebraic expression can be attached to the wrong data geometry.

Contents · Previous chapter · Next chapter

For Q1 = 4 and Q2 = 6 mmol/g with Δ = one minute, the remaining fraction is r = 6/4 − 1 = one half. The rate coefficient is therefore ln 2 per minute. The endpoint is 4/(1 − 1/2) = 8 mmol/g.

Check through the alternate endpoint expression: Q1²/(2Q1 − Q2) = 16/(8 − 6) = eight. Then reconstruct both observations from q(t) = 8[1 − e^(−t ln 2)]. The inverse calculation should return the original record rather than only produce plausible-looking parameter values.

Do the same with specimen B’s first two readings, four and 20/3. Their ratio minus one is two thirds, and the recovered endpoint is twelve. The extra reading has distinguished the two specific candidates. That success remains conditional on the shared model, initial condition and observation timing. It does not establish that two measurements can validate the full physical mechanism of a real zeolite specimen.

Contents · Previous chapter · Next chapter

Positive finite k implies 0 < r < 1. Consequently an increasing two-reading record must satisfy Q1 < Q2 < 2Q1. The upper inequality expresses the slowing approach: the second interval’s increase is smaller than the first interval’s increase from zero.

If Q2 equals Q1 exactly, the inverse gives r = 0, which is not produced by a finite positive k in this model. If Q2 equals 2Q1, it gives r = 1 and no finite qe compatible with positive uptake from zero. A value beyond either side should not be repaired by blindly taking a logarithm of an inadmissible number.

These are exact compatibility statements for exact teaching values. Real noisy measurements require an uncertainty-aware analysis; they should not be rejected mechanically because rounding nudges a point across a boundary. The practical habit is to investigate the data, timing, baseline and model together. Mathematical incompatibility is a question to resolve, not permission to hide an inconvenient observation or substitute a more flattering measurement.

Contents · Previous chapter · Next chapter

Two admissible readings determine qe and k under our assumptions, leaving no independent point with which to check the fitted shape. A third equally spaced reading supplies a useful prediction test. The successive increments must have the same ratio r.

For specimen A, the first increment is four, the second is two and the predicted third is one mmol/g. Therefore Q3 should be seven. The formula is Q3 = Q2 + r(Q2 − Q1), using r inferred from the first two observations.

If the supplied third reading were 7.4, the exact two-parameter curve inferred from four and six would not reproduce it. That discrepancy could reflect measurement uncertainty, timing error or inadequacy of the model. The three numbers do not identify which explanation is correct. Record the mismatch rather than silently refitting different pairs and reporting whichever endpoint is most convenient. A predictive check is useful precisely because it can reveal something the initial fit was forced to overlook.

Contents · Previous chapter · Next chapter

The time constant τ is 1/k. It is the time over which the remaining gap is multiplied by e^(−1), approximately 0.368. The half-time for the remaining gap is t½ = ln 2/k. Those times are related, but they are not numerically equal.

For specimen A, k = ln 2 per minute, so the half-time is one minute and the time constant is approximately 1.443 minutes. Calling one minute its time constant would mislabel the mathematical relationship even though one minute is a useful observation time.

A target such as 90% of the equilibrium loading gives another time: t90 = ln 10/k. For A this is approximately 3.322 minutes. It refers to a fraction of the stipulated endpoint, not 90% removal from an unspecified fluid. A percentage of loading and a percentage of fluid removal have different denominators. Keep the target definition attached to the time so a reader can reproduce the calculation without inventing a vessel inventory or flow rate.

Contents · Previous chapter · Next chapter

Suppose the task is to reach four mmol/g. Both fictional specimens attain it after one minute, despite different endpoints and coefficients. For another target below each endpoint, solve t = −ln(1 − qTarget/qe)/k separately for each specimen.

The target must lie between the initial loading and qe for a positive finite crossing time in this model. A request to reach nine mmol/g is impossible for specimen A’s endpoint of eight, while it is attainable for specimen B’s endpoint of twelve. A faster early approach does not overcome an insufficient endpoint.

This is why the word faster needs a reference. One can compare coefficients, half-times, times to a common fractional endpoint, or times to a common absolute loading. Those comparisons answer different questions and need not produce one universal ranking. State the application target before choosing the metric. The calculation can clarify that target while leaving actual material selection, operating conditions and whole-system performance to the relevant evidence and qualified assessment.

Contents · Previous chapter · Next chapter

At the zero start, the initial slope is k qe. It combines endpoint and coefficient. A large initial slope can come from a large endpoint, a large coefficient, or both. Measuring that slope alone therefore does not generally identify the two parameters separately.

Our first specimens have initial slopes 8 ln 2 and 12 ln(3/2) mmol per gram per minute. Their first-minute loadings coincide even though those initial slopes differ. The curve’s shape over the interval matters; an endpoint difference divided by the interval is an average rate, not automatically the instantaneous starting slope.

When interpreting a short record, ask whether the instrument actually resolves the early transient and whether time zero is known. A fitted straight line over several samples may estimate a local or average change under its own assumptions. It should not be relabelled as the exact derivative of the ideal model without checking the approximation. Preserving the distinction helps students move between graphs, slopes and differential equations without treating every rate number as the same quantity.

Contents · Previous chapter · Next chapter

For small kt, the exponential expansion gives q(t) approximately qe kt. To first order, the early loading depends mainly on the product qe k. Several endpoint-and-rate combinations can then produce similar early traces, even when more than one observation is available.

This is a conditioning issue rather than the exact one-point ambiguity alone. The full nonlinear model may mathematically distinguish two parameters from exact readings, while small errors in short-time measurements make the distinction unstable. More decimal places in the fitted endpoint do not automatically represent more information.

For the equal-spacing inverse, the warning appears in the denominator 2Q1 − Q2. If the record is almost linear from the zero start, Q2 is near 2Q1 and that denominator is small. The inferred endpoint can become large and sensitive. A useful evidence plan considers observation times that reveal curvature, while respecting the actual experimental constraints. In a paper exercise, students can test the sensitivity simply by changing the supplied second reading and recomputing the endpoint.

Contents · Previous chapter · Next chapter

Keep Q1 = 4 mmol/g, but compare Q2 = 7.80 with Q2 = 7.90. Both satisfy the exact compatibility inequalities. The inferred endpoint is 16/(8 − 7.80) = 80 mmol/g in the first case and 16/(8 − 7.90) = 160 mmol/g in the second.

A change of 0.10 mmol/g in the second reading doubled the inferred endpoint. The corresponding remaining fractions are 0.95 and 0.975, giving different positive rate coefficients. These numbers are fictional sensitivity examples, not plausible-capacity claims for a named material.

The calculation also explains why a known physical constraint can be useful without manufacturing a unique answer. If independent evidence limited qe, some candidate pairs would become inconsistent with the simple model and that evidence together. The proper response would be to examine the combined assumptions and uncertainties. It would not be to clip the endpoint to a convenient limit and retain the same claim of exact identification. Constraints and data must describe one coherent account.

Contents · Previous chapter · Next chapter

If the specimen starts at a known loading q0, the solution becomes q(t) = qe + (q0 − qe)e^(−kt). The changing quantity is still the gap to qe, but its initial size is qe − q0 rather than qe. Subtracting q0 from each reading can recover a similar approach-from-baseline form when the relevant conditions hold.

If q0 is unknown, the two-reading formulas derived for a known zero start no longer identify all the unknowns. The model now contains an additional baseline quantity. Treating the first recorded point as zero loading would silently replace the initial-condition question with an assumption.

This matters in repeated use. A regenerated specimen may retain some adsorbate, and a new measurement may begin from that retained amount. A smaller change during the next cycle does not alone establish that the equilibrium endpoint fell. The starting state can also have changed. Keep the measured state, the difference from the start and the equilibrium target separate before attributing a difference to ageing, damage or another physical cause.

Contents · Previous chapter · Next chapter

The earlier two-reading inversion needed q0 = 0. With an unknown q0, three equally spaced nonzero-time readings can provide another relationship under the same constant-endpoint model. Their successive increments have ratio r = (Q3 − Q2)/(Q2 − Q1), provided the denominator is nonzero and the result lies in the admissible range.

For a concrete comparison, retain Q1 = 3 and Q2 = 4.5 mmol/g at two and four minutes. One compatible model has q0 = 0, qe = 6 and a two-minute remaining fraction of one half. It predicts Q3 = 5.25. Another has q0 = 1, qe = 9 and a two-minute remaining fraction of three quarters. It reproduces the same first two observations, but predicts Q3 = 5.625 at six minutes.

The latter check is explicit: nine minus eight times three quarters gives three; nine minus eight times nine sixteenths gives 4.5. The third point distinguishes these two exact models. Once r is determined from the increments, qe can be recovered from (Q2 − rQ1)/(1 − r), and the initial state can then be reconstructed from Q1 = qe + (q0 − qe)r. This remains a conditional algebraic identification, not proof that the measurement process or physical mechanism follows the equation.

Contents · Previous chapter · Next chapter

Suppose the actual adsorption starts before the reported clock begins. A curve recorded from that later time can resemble a curve with a nonzero initial loading. The time-origin shift changes what the first record means, even if the subsequent evolution follows the same ideal coefficient.

For a constant-qe exponential, delaying the observation origin by h multiplies the remaining-gap amplitude by e^(−kh). The coefficient k itself stays unchanged. If an analyst instead forces the delayed record through an invented zero point, the recovered endpoint or coefficient may be distorted.

This is an observation-contract problem. The record should state what event defines time zero and how it is connected to the measurement. A timestamp alone does not establish that contact began at that instant. The paper model makes the issue visible without prescribing experimental hardware. Before comparing apparent rates across laboratories or specimens, check whether their timing definitions, initial conditions and response windows describe the same stage of the process.

Contents · Previous chapter · Next chapter

Our simple solution assumed constant qe. If the surrounding gas composition or pressure changes, the relevant equilibrium loading may become qe(t). The equation dq/dt = k[qe(t) − q] can still be written as a chosen model, but its solution is not generally the same constant-endpoint exponential.

For constant k, the solution includes the initial term q0 e^(−kt) plus an integral of k qe(s)e^(−k(t − s)) over the earlier history. This expression says that the current state depends on the past forcing, not just the present value of qe. No integral evaluation is needed to see why replacing a changing input with its final value can misdescribe an entire uptake record.

A sealed finite reservoir can introduce a related coupling: uptake changes the fluid state, which changes the equilibrium target. That requires mass balances and an appropriate equilibrium relationship. The maintained-environment exercise deliberately excluded this feedback. Recognising the difference is a practical transfer check. A formula should travel to a new setting only when the assumptions that produced it travel as well.

Contents · Previous chapter · Next chapter

An exponential approach can arise in several simplified dynamic accounts. A fitted k is therefore a parameter of a stated model and observation basis, not automatically a microscopic diffusion coefficient or proof of one transport mechanism. Establishing a physical mechanism requires additional evidence suited to that question.

Likewise, an excellent fit over a selected interval may conceal inadequacy elsewhere. Check predictions at unused times, examine residual patterns and preserve the conditions under which the record was obtained. Adding more parameters can improve an in-sample fit without establishing a better transferable explanation.

The primary study by Ryu and colleagues on water adsorption in zeolite 13X compared equilibrium and kinetic models for its particular measured material. Its journal abstract reports different performance among model alternatives. That research context supports taking model choice seriously; it does not validate our fictional parameters or establish that every zeolite follows the same constant-endpoint law. Read the original journal record with its material and conditions attached. Source route checked on 24 October 2026.

Contents · Previous chapter · Next chapter

The variable q describes loading on the stipulated solid basis. A breakthrough curve describes how an outlet concentration changes during a flowing experiment. Connecting them requires an account of fluid transport, inventories, flow, spatial variation and any other relevant processes within the chosen model.

A local uptake coefficient does not by itself supply a column service time. Neither does a final equilibrium loading identify when an outlet will cross a particular concentration criterion. Those questions can be related, but the relationship has to be modelled rather than assumed from a familiar exponential shape.

For learners, draw two separate boxes: one for the solid loading record and one for the fluid entering and leaving a defined system. Label which quantities are measured in each. Then ask which balances and relationships connect the boxes. The Zeolites root’s changing-outlet example develops a cumulative material account, while this lesson develops parameter identification for a stipulated loading-time curve. Preserving their different observables makes the connection useful without merging their guarantees.

Contents · Previous chapter · Next chapter

Consider the statement: both specimens load four mmol/g after one minute, so they have the same capacity and rate. The observation establishes one equality of time-specific loading on the supplied basis. It establishes neither of the two claimed parameter equalities.

Use specimens A and B to repair the explanation. They reproduce the same first observation, while their endpoints are eight and twelve and their coefficients are ln 2 and ln(3/2). The next useful observation should address the unresolved curve, rather than merely repeat the first loading with a more impressive number of decimal places.

A responsible revised statement might say that the specimens match at the observed time under the recorded conditions, while their equilibrium loadings and approach coefficients remain unresolved by that observation alone. If a second admissible reading is supplied and the model assumptions are maintained, the two parameters can be inferred conditionally and checked with another time point. The correction becomes a learning tool when it explains which new information changes the conclusion and which assumptions still carry the inference.

Contents · Previous chapter · Next chapter

For a new fictional specimen, suppose q(0) = 0, q(2 minutes) = 3 mmol/g and q(4 minutes) = 4.5 mmol/g. Assume the same constant-endpoint linear-driving-force model. Determine the remaining-gap factor over two minutes, the equilibrium loading and k in per-minute units.

Predict the loading after six minutes. Then calculate the remaining-gap half-time and the time constant, keeping their definitions separate. Finally ask whether the same observations would identify all parameters if the initial loading were unknown. Explain the additional ambiguity without inventing an initial state.

Start from the paired equations rather than copying a coefficient from the earlier one-minute example. The interval is now two minutes. A correct remaining fraction with the wrong time conversion would produce an incorrect k while leaving some intermediate arithmetic apparently familiar. Check the answer by reconstructing both supplied observations and by comparing the predicted increments. The transfer is successful when the changed timing becomes part of the reasoning automatically.

Contents · Previous chapter · Next chapter

Here Δ = two minutes and r = 4.5/3 − 1 = one half. The endpoint is qe = 3/(1 − 1/2) = 6 mmol/g. The coefficient is k = ln 2/2 per minute, approximately 0.3466 per minute.

At six minutes, the remaining gap is six times (one half) cubed, or 0.75 mmol/g. The predicted loading is 5.25 mmol/g. The successive increases over the two-minute intervals are three, 1.5 and 0.75, confirming the common ratio.

The half-time is two minutes, while the time constant is 2/ln 2 minutes, approximately 2.885. If q0 were unknown, the two nonzero-time observations would not identify q0, qe and k uniquely. The zero-start condition is doing real work in the inverse calculation. State it alongside the inferred values so another reader does not mistake a conditional teaching solution for a complete parameter determination from an unspecified experimental record.

Contents · Previous chapter · Next chapter

If the endpoint is independently known, a positive-time reading below that endpoint can help infer k under the model. If the endpoint is unknown, appropriately timed readings that reveal curvature can be more informative than a single early point. If the initial condition is uncertain, evidence about the start may be more valuable than another copy of the same summary.

Measurement uncertainty belongs in that plan. A late record near the endpoint can be sensitive when estimating a small remaining gap, while an early almost-linear record can be sensitive when estimating the endpoint. There is no universally best observation time independent of the target, prior information and measurement process.

For classroom practice, compare complete fictional candidate curves and ask which additional reading would distinguish them. Then remove an assumption and state what becomes unresolved. This turns a fitted parameter into an inspectable inference. The useful habit is to choose evidence for the missing relationship, rather than collect more data solely because the existing graph looks smooth or because a particular sampling interval is familiar.

Contents · Previous chapter · Next chapter

Is the largest measured loading automatically the equilibrium capacity?

No. It is the largest observed loading in the supplied record. Calling it an equilibrium endpoint needs an appropriate criterion and evidence. Our exact exponential approaches its endpoint without reaching it at a finite time.

Does a larger k always make a specimen better?

It makes the fractional approach faster within the stated constant-endpoint model. A practical comparison also needs the endpoint, initial state, target and system conditions. A rapid approach to an insufficient loading does not meet a larger loading requirement.

Can two readings prove the mechanism?

They can identify two parameters conditionally under the specified timing, baseline and model. They cannot independently validate every physical assumption used to produce that result. Additional observations and relevant material evidence are needed to challenge the account.

Is the approach coefficient a diffusion coefficient?

Not automatically. Its units here are inverse time. Relating it to a microscopic transport parameter requires an additional model with its geometry and assumptions stated. Copying a familiar parameter name does not supply that relationship.

Contents · Previous chapter · Next chapter

For repair, practise identifying loading, gap and rate on a labelled curve. For stabilisation, reconstruct the two-reading inversion and verify both observations by substitution. For extension, add a third predictive check, a nonzero initial condition or a changing equilibrium target and explain which earlier formula no longer applies.

Return to the Zeolites root for material identity, timing and complete treatment boundaries. Read the zeolite science guide for the wider distinction between equilibrium, transport and experimental evidence. Use the finite-bath ion-exchange branch when conservation and equilibrium must be solved together for that different mechanism.

For a different equilibrium question, the competitive-uptake lesson keeps shared-site mixture occupancy separate from the time required to approach a specified state.

For a related inverse problem, the pump-down record lesson distinguishes an indicated pressure floor from physical gas load and gauge offset. Its pressure observable differs from the loading model here; the connection is the need to preserve the observation and baseline assumptions when interpreting an exponential.

Continue through Learning That Transfers into Life and How X Works. A useful final explanation states the observed quantity, initial condition, chosen model, identified parameters and remaining limits. The learner should be able to explain why the next observation matters, rather than only reproduce the fitted curve.

Contents · Previous chapter · Return to Zeolites