Open Access
EPJ Nuclear Sci. Technol.
Volume 9, 2023
Article Number 28
Number of page(s) 6
Published online 24 October 2023

© B. Mercier and V. Borysenko, Published by EDP Sciences, 2023

Licence Creative CommonsThis is an Open Access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

1. Introduction

On the web, we find which states:

The response of a reactor to a change in temperature (i.e., the overall reactor stability) depends especially on the algebraic sign of αT. A reactor with negative αT is inherently stable to changes in its temperature and thermal power, while a reactor with positive αT is inherently unstable.”

In the present paper, we consider the case of a reactor for which αT is positive at low power and negative at large power. This case is of interest since this was how the RBMK behaved at the time of the Chernobyl accident.

Let k be the neutron effective multiplication factor.

Let ρ = k − 1/k denote the reactivity.

Obviously when ρ > 0 (resp. ρ < 0 the reactor is supercritical (resp. subcritical). In other words, the linear theory predicts that the reactor is unstable. It also tells us that when the reactor is critical (i.e. ρ = 0) to increase its power, one has to make a positive reactivity injection ρ0 > 0, e.g. by moving up the control rods, and then, when the power has reached the desired power level, a negative reactivity injection −ρ0 e.g. by moving down the control rods where they were before.

This is not, as we shall see, how it works in a real nuclear reactor where there is a reactivity feedback.

Let n denote the normalized reactor power such that n = 1 at nominal power, the linear theory predicts that


where τ is the effective neutron lifetime, which means that n(t) increases or decreases exponentially.

In a PWR, when the power increases, the fuel temperature increases inducing the Doppler effect. The moderator temperature also increases, inducing a moderator effect.

In other words, there is a negative reactivity feedback α(n) < 0 such that


Therefore ρ is not a constant and we should replace (1) by


Note that a point reactor model is used for n; τ takes into account the delayed neutrons; α takes into account both moderator and Doppler effects and is supposed to depend only on n.

In view of

we obtain


Finally, we have to study the 2 × 2 nonlinear differential system:

(5) (6)

Note that we can exclude the particular case n0 = 0 which is not of interest.

2. Reactor stability

If, for a while, we consider ρ as a function of n, we have the differential equation (2) which should be complemented by the “initial” condition


So that we have


which holds both for n > n0 and for n < n0.

The reactivity injection ρ0 can be positive (in which case the reactor power increases) or negative (in which case it decreases).

The reactor stability requires, as we shall see, that there exists α* such that:


Indeed, with such an assumption, for all starting point {n0, ρ0}, the curve:

will cut the horizontal axis once and only once in the plane {n, ρ}. In other words, the equation


has a unique solution. Note, however, that we may have n < 0.

The particular case where α(n) is constant has been considered in Reuss-Mercier [1].

In such a case, we have

so that the curve nρ = ρ(n0, ρ0, n) is a straight line.

In this section, we consider the general case described by (9).

Proposition 1: Under the above assumption (9), then n(t) → max(0, n) when t → ∞.

Proof. We divide (4) by ρ and (3) by n so that we get

(11) (12)

1°/Case where ρ0 > 0.

Equation (12) shows that n(t) increases, at least in the neighborhood of t = 0, but can reach a maximum only if ρ vanishes which means n(t) = n.

As long as n(t) < n, we have ρ(t) = ρ(n0, ρ0,n(t)) > 0.

Since we have

we deduce that Logρ(t) → −∞ if t → ∞ so that ρ(t) → 0.

The curve nρ = ρ(n0, ρ0, n) being continuous, we deduce that n(t)n.

2°/Case where ρ0 < 0.

In an analogous way, (12) shows that n(t) must be decreasing, at least in the neighborhood of t = 0 but can reach a minimum only if ρ vanishes, then if n = n.

Let us assume first that n > 0.

We have nn(t)no and then α(n(t)).n(t) < α*.n(t) < α*.n < 0.

Therefore Log|ρ(t)| → −∞ and then |ρ(t)| → 0 so that n(t)n.

Let us now assume that n < 0.

The curve nρ(n0, ρ0, n) cuts the vertical axis n = 0 in ρ = ρ* < 0.

We deduce that ρ(t) increases and that ρ(t)ρ*.

Since ρ0ρρ* < 0 we deduce from (12) that Log n(t) → −∞ and then n(t) → 0.

With this result, we have shown that for a stable reactor, due to the reactivity feedback, to increase the power of the reactor, it is sufficient to make an appropriate reactivity injection ρ0, but it is not necessary to make a negative reactivity injection equal to −ρ0 after this positive reactivity injection.

3. Evaluation of α(n) for the RBMK

It is well known that, before the Chernobyl accident, there was a significant void effect on the RBMK. This void effect originated from the fact that graphite is used as a moderator and water as a coolant. Since light water absorbs thermal neutrons, if it disappears, the core reactivity increases.

More precisely, in [2, p.55] we find that the Doppler effect brings a negative reactivity of −1000 pcm and that the void coefficient may be as high as +2500 pcm. (This was the case just before the accident).

We assume that the Doppler effect is proportional to n(e.g. −100 pcm for n = 0.1.

Table 1.

Contributions of the coolant and Doppler effects to reactivity.

To evaluate the coolant effect, we assume that

  • the coolant pressure is 7 MPa, then the saturation temperature is 285.8 °C.

  • As for the enthalpy, we have hl,sat = 1267 kJ/kg for saturated liquid hv,sat = 2772 kJ/kg for saturated steam.

  • With obvious notations, we have ρl,sat = 739 kg/m3 and ρv,sat = 36 kg/m3.

  • The associated specific volumes are τl,sat = 1.35 L/kg and τv,sat = 27.8 L/kg.

  • The mass flow rate is q = 10417 kg/s.

The inlet enthalpy is adjusted in such a way that the saturation level is at zsat = 1.855 m (note that H = 7 m is the height of the core).

To obtain the values given in Table 1, we proceed in the following way.

Given the reactor power P, we calculate the enthalpy increment per meter Δh = P/(H.q).

We calculate hin = hl,sat − zsat * Δh and hout = hl,sat + (Hzsat) * Δh.

Then we calculate the steam mass fraction

xout = (houthl,sat) /(hv,sathl,sat), τout = (1 − xout) τl + xoutτv and ρout = 1000/τout.

We assume that ρ(z) is constant for 0 ≤ zzsat and then varies linearly from ρl,sat for z = zsat to ρout for z = H.

We calculate its average ρave in the pressure tubes and then the coolant effect (in pcm).

Coolant effect = 2500 * (ρl,satρave) /ρl,sat

(In the limit case ρave = 0 we would get that the coolant effect would be equal to the void effect).

Starting from criticality at n = 0, the core reactivity for n ≠ 0 is computed as the algebraic sum of the (positive) coolant effect and the (negative) Doppler effect. It is computed in column 6 of Table 1. The derivative in pcm/MW is computed in column 7 and plotted in Figure 1.

thumbnail Fig. 1.

nα(n) for the RBMK before the accident.

Finally, our α(n) would be obtained by multiplication of column 7 by 3200.

What it means is that rather than being always negative, α(n) is positive for n < n* and α(n) < 0 when n > n* with n* = 0.238 that is P = 762 MW.

Clearly, condition (9) is violated.

In Section 4, we shall prove that, in such a case, the reactor is unstable and we shall study how it is unstable.

4. Behavior of the unstable reactor

In this section, we shall assume that there exists n* such that

0 < n* < 1 and α(n*) = 0

α(n) > 0 for n < n*

α(n) < 0 when n > n*.

The curve nρ = ρ (n0, ρ0, n) that we have defined in (8) is then increasing when n < n* and strictly decreasing when n > n* as indicated in Figure 2. Equation (10) may then have 2 solutions: one smaller than n* and the other one greater. We shall prove that only the latter (which we denote by n) is stable. We shall prove that

  • if ρ0 > 0 then n(t)n when t → ∞,

  • if ρ0 < 0 and n0 > n*, then n(t)n when t → ∞,

  • if ρ0 < 0 and n0 < n*, then n(t) → 0 when t → ∞.

thumbnail Fig. 2.

Example of curve nρ = ρ (n0, ρ0, n).

Let us first consider the case where ρ0 > 0.

Let us choose n1 > n* such that ρ(n0, ρ0, n1) ≥ ρ0 (see Fig. 2) and then

From (12) we deduce that

and then, as long as n(t)n1

Then the set of real numbers t such that n(t)n1 is bounded. Then, there exists a time t1 such that n(t1) = n1.

Let ρ1 = ρ(n0, ρ0, n1); we have ρ(t1) = ρ1.

Let α* = min {ρ(n):n > n1} with the above assumptions, we have α* < 0.

Since α(n) < α* < 0Ɐn > n1 we can apply proposition 1 which tells us that n(t)n.

The case where ρ0 < 0 and n0 < n* is illustrated in Figure 3.

thumbnail Fig. 3.

Curve nρ = ρ (n0, ρ0, n) obtained for n0 = 0.05 and ρ0 = −26.6 pcm.

Equation (12) shows that n decreases and hence ρ also decreases. We see that Log n → −∞ hence n → 0.

To be complete we should also consider the case where n0 > n* and ρ0 < 0.

We can just apply proposition 1, which shows that ρ(t) → 0 and n(t)n.

We have thus proved the desired result.

Note that we have proved that the starting points n0 < n* with ρ0 = 0 are unstable:

  • if rather than being equal to zero, ρ0 is slightly positive we shall have n(t)n > n*.

  • If ρ0 is slightly negative we shall have n(t) → 0.

Note also that with 3 starting points n01 = 0.05, n02 = 0.12 and n03 = 0.5 and the same reactivity injection ρ0 = 25 pcm, as can be seen in Figure 4 the final power n may be larger when the initial power n0 is smaller.

thumbnail Fig. 4.

3 starting points n01, n02, n03 with the same ρ0 = 25 pcm.

5. Conclusion

With our simple model, we have shown that reactivity feedback is an important phenomenon.

When the reactor is stable, to increase its power, it is sufficient to make an appropriate reactivity injection ρ0, but, as the operators know it, it is not necessary to make a negative reactivity injection equal to −ρ0 after this positive reactivity injection.

We have also studied one case of a reactor that is unstable at low power, like the RBMK.

We have shown the practical effects of such an instability:

  • 1)

    with the same reactivity injection, the lower the initial power then the final power may be higher,

  • 2)

    the line ρ0 = 0 0 < n0 < n* is a bifurcation line.

  • If we start from ρ0 slightly positive we have n(t)n > n*;

  • if we start from ρ0 slightly negative we have n(t) → 0.

This was the case for the RBMK before the Chernobyl accident: the operators were obliged to frequently use the control rods, and, practically, remaining at low power was not comfortable at all for them. This is one explanation for the accident, but of course, it is not the only one [3, 4].

Note that our 2 × 2 model can only give a qualitative behavior of the reactor; to get accurate kinematics, we would have to adjust the effective neutron lifetime τ or use the 8 × 8 model with delayed neutrons, as described in [1].

As has been shown in [1], the asymptotic behavior is the same for both models.

Our results explain better why Anatoly Dyatlov has written in

The nuclear safety department of the power plant (…) measured the fast power coefficient at power levels close to nominal full power, ie in the region where it was negative. The results were used by the operators in their everyday work. The latest data prior to the accident gave a value of minus 1.7 × 10−4 ß/MW”.

Conflict of interests

The authors declare that they have no competing interests to report.


This research did not receive any specific funding.

Data availability statement

This article has no associated data generated and/or analyzed.

Author contribution statement

1. Introduction B. Mercier – 2. Reactor stability B. Mercier – 3. Evaluation of α(n) for the RBMK V. Borysenko – 4. Behavior of the unstable reactor. B. Mercier.


  1. P. Reuss, B. Mercier, A simple model to show the effect of counter-reactions, EPJ Nuclear Sci. Technol. 9, 16 (2023). [CrossRef] [EDP Sciences] [Google Scholar]
  2. B. Barré, P. Anzieu, R. Lenain, J.B. Thomas, Nuclear Reactor Systems (EDP-Sciences, 2016) [Google Scholar]
  3. V. Borysenko, V. Goranchuk, A reactivity accident thirty- six years later, in International Conference on Nuclear Decommissioning and Environment Recovery, INUDECO 2022 27–28 April, 2022, Slavutych (Ukraine) [Google Scholar]
  4. GRS-121, The accident and the Safety of RBMK-Reactors, 1996 [Google Scholar]

Cite this article as: B. Mercier, and V. Borysenko. Technical note: stable and unstable reactors, EPJ Nuclear Sci. Technol. 9, 28 (2023)

All Tables

Table 1.

Contributions of the coolant and Doppler effects to reactivity.

All Figures

thumbnail Fig. 1.

nα(n) for the RBMK before the accident.

In the text
thumbnail Fig. 2.

Example of curve nρ = ρ (n0, ρ0, n).

In the text
thumbnail Fig. 3.

Curve nρ = ρ (n0, ρ0, n) obtained for n0 = 0.05 and ρ0 = −26.6 pcm.

In the text
thumbnail Fig. 4.

3 starting points n01, n02, n03 with the same ρ0 = 25 pcm.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.