In mediaeval times, alchemists spent their days searching for a way to change metals such as mercury into gold. They were looking for the philosopher’s stone; a way of getting something for nothing (Britannica 2024). They were unsuccessful because they didn’t understand the nature of matter. Modern scientists would not make such foolish mistakes. Or would they?

Consider the search for phosphorus solubilizing microorganisms. The background has been explained in detail many times. Briefly, phosphorus (P) is an essential element for life. It is thought to be an unavoidable cost for many food producing systems. World reserves of high-quality phosphate ore are limited (Johnston et al. 2014). Farming practice in many parts of the developed world involves applying more phosphate than will be removed in the harvested crop. Consequently, large amounts of phosphate have accumulated in many soils. For example, Ma et al. (2018) calculated that in 2012 the difference between P input and P output for arable soils in China was 42.1 kg P ha−1. Schoumans and Chardon (2015) reported that the average P accumulation in agricultural soils in the Netherlands was about 2050 kg ha−1. This phosphate is known as legacy P. If we could find an easy and cheap way to utilise it, this would be something for nothing. Sound familiar?

Many ways of seeking access to this phosphate have been suggested. Some of these involve utilising strategies plants have evolved. However, these involve something for something. Many plants trade the products of photosynthesis with mycorrhizal fungi and take mineral nutrients, especially phosphate, in exchange. Mostly, fungal hyphae seem to act as super-thin roots going to where the P is and decreasing the distance it must diffuse through the soil solution (Bolan 1991). Other plants have adopted a different strategy; they secrete organic acids often utilising specialised structures called cluster roots (Shane and Lambers 2005). Again, this is something for something.

The true philosopher’s stone, the one that gets you something for nothing, is the use of phosphate solubilizing microorganisms. The search for P solubilizing microorganisms began 76 years ago (Pikovskaya 1948). (The elapsed time suggests that application of this idea may not be quite straightforward.) This work set a pathway that has been followed by many others; a Web of Science search in May 2024 using the search words “phosphate solubilizing microorganisms” returned 2014. hits. This may be an underestimate of the true activity because of variations in spelling (“solubilising” for example) and also perhaps because “unsuccessful” experiments, ones that don’t give a positive response, are under-reported.

Typically, a sample of soil is taken, often from the rhizosphere, and its microorganisms are offered a medium containing glucose and inorganic salts including an ammonium salt. Organisms that can grow on, and acidify, this medium are selected. Acidification is usually detected by the dissolution of a sparingly soluble P compound such as tricalcium phosphate, but bromophenol blue, an indicator, has also been used (Gupta et al. 1994; Mehta and Nautiyal 2001). That is, it is the acidity that matters. The chosen microorganisms are then cultured and returned to a soil.

Consider the assumptions of this process. The first assumption is often that soil phosphate is present as discrete phosphate compounds. Kishore et al. (2015), in a paper they regard as a critical review, state that soil phosphate is present as strengite, variscite, and several calcium compounds. Similar assumptions are expressed or implied in many papers. One can hardly be surprised at this; it is what is taught in most textbooks. However, it is not in accord with easily made observations. There are many published smooth curves of phosphate sorption against concentration. If precipitation were important this would not be possible. Once some solubility product is exceeded, apparent sorption would increase with little increase in concentration because of precipitation. Further arguments against the formation of such discrete compounds as a result of fertiliser application are given by Barrow (2020). For example, the precipitation theory cannot explain why increasing salt concentration increases sorption at high pH but decreases it at low pH. This can be explained by assuming that the initial reaction is with variable charge surfaces. Further, formation of discrete compounds would be a reversal of soil formation and would involve a decrease in entropy. Phosphate behaviour is much better described by assuming that it is mostly present in soil as adsorbed and penetrated on variable charge surfaces. The assumption that phosphate is present in soil as discrete compounds is an important foundation stone in theories about phosphate solubilizing microorganisms.

It has been suggested to me that discrete phosphate compounds might be present originating from the parent rock rather than newly formed. This is not my local experience because I live in an ancient corner of an ancient continent but might be possible in younger soils. However, any such compounds must be very resistant to attack for otherwise they would not have persisted.

The second assumption is related to the first. It is that acidifying a soil will improve access to soil phosphate. Strangely, this assumption was made even though the classical belief is that the optimum pH for phosphate uptake is 7 (Penn and Camberato 2019). If this were true, acidification could only be effective if the original pH was higher than this. It is even stranger that there have been few measurements of the optimum pH for phosphate uptake. The few that are available suggest that this optimum is about pHCaCl2 5 (Barrow and Hartemink 2023). Above pHCaCl2 5 acidification will improve utilisation of phosphate, but the effect is mainly via the effects of pH on the physiology of plant roots (Barrow and Hartemink 2023). Many of the reported beneficial effects of acidifying microorganisms are via this mechanism (Barrow and Lambers 2022). Below pHCaCl2 5 acidification would be harmful. In their review, Raymond et al. (2021) cite two examples. In one, plant response occurred in alkaline soil but not in an acid one. In another, inoculation was more effective in an alkaline soil than in a neutral one.

The third assumption is that organisms that are normally a minor constituent of the soil biota, and which respond to a ready source of energy such as glucose, can become a major constituent if they are returned in large numbers to a soil in which many other microorganisms are competing for any available source of energy. Raymond et al. (2021) concluded that most positive responses to inoculation come under controlled conditions whereas responses under field conditions are rare.

The fourth assumption is that microorganisms, in a situation in which growth is probably limited by energy supplies, benevolently donate energy-rich compounds, such as organic anions, for purposes that do not benefit them directly.

Nevertheless, there are many reports of beneficial responses from the use of phosphate solubilizing microorganisms. Some responses have been attributed to stimulation of root growth (for a list of citations see Raymond et al. (2021)). Stimulated root growth might gather more P, but this is something for something. A similar effect could be obtained by varietal selection. For some, enthusiasm for the concept has overshadowed normal scientific scepticism. Alori et al. (2017) include a table with 11 entries of the beneficial effects of some phosphate solubilizing microorganisms on crops. Some of these involve nitrogen fixing organisms rather than phosphate solubilizing organisms. One reports increased P uptake, nutrient content, and fresh weight of oil palm, but the original publication (Istina et al. 2015) notes that the differences were not significant. Some of them involve treatments with phosphate rock or tricalcium phosphate. This is also a common theme – for example (Panhwar et al. 2011). I accept that organisms selected on their ability to dissolve sparingly soluble phosphate compounds are likely to be able to dissolve them near plant roots. That they should make rock phosphate more available is not surprising, but I suggest that there are few situations in which this is relevant. Of the 11 examples of Alori et al. (2017) only one (Jahan et al. 2013) shows a clear response. It was for improved growth of sesame in a soil of pH 7.8. It is therefore explainable in terms of effects of decreasing pH on the rate of uptake of phosphate by plant roots. How can we distinguish this effect from effects on solubilization of soil phosphate?

Consider the interpretation of responses. When a plant responds to applied phosphate, the response can be described by:

$$y=m (1-\text{exp}\left(-c \left(x+d\right)\right)$$
(1)

where y is some measure of the response such as dry weight, and x is the amount of phosphate applied. The parameter m indicates the maximum to which y trends when fertiliser application is large; the parameter c indicates the slope of the response curve, the greater its value the less phosphate is required for a given yield; parameter d indicates the phosphate coming from the soil and the seed.

When a treatment, such as P solubilizing microorganisms, is applied, and it increases the value of y, it does not necessarily follow that this is caused by an increase in the value of d and that the treatment has released (solubilized) phosphate from soil. It could be due to increase in the other parameters; a move to a more favourable pH would be reflected in an increased value of the parameter c; removal of some kind of restriction, such as control of pathogens, would be reflected in an increased value of the parameter m. Thus, increased production at one level of phosphate does not necessarily indicate phosphate solubilization. A similar argument has been put forward by Raymond et al. (2021). To prove that phosphate solubilizing has occurred it is necessary to measure phosphate response curves in the presence and the absence of such treatments. There are very few such experiments.

Is there any alternative, any way to avoid the endless application of large amounts of phosphate? Perhaps there is.

Consider the way that phosphate reacts with soil particles. An initial adsorption reaction on variable charge surfaces is followed by diffusive penetration (Barrow 2020). This increases the negative charge on the reacting surfaces, and this decreases their propensity to adsorb further phosphate. The distribution of phosphate between the solid and liquid phase moves towards liquid phase. Plots of phosphate sorption against concentration have a lower slope (Fig. 1a); the buffering capacity is thus decreased. This increases the rate of diffusion to plant roots and means that further applications of phosphate are more effective (Fig. 1b). The diffusive penetration does not continue for ever. We can think of the penetrated phosphate as “repairing” pores. As a result, it becomes more difficult to extract both iron and aluminium (Fig. 2). Sorption and desorption then tend to have a similar slope; hysteresis disappears (Figs. 3b and 4). A major sink for phosphate, and one of the most important reasons for repeated applications, then disappears. Effectiveness of applied phosphate no longer decreases with increasing time of contact (Fig. 3d). These effects are summarised in Fig. 4. It shows that the decreased buffering capacity decreased P requirement to one quarter of that of the original soil. Further, any application would retain its effectiveness rather than decreasing with increasing time of contact (Fig. 4b.) These effects can be measured in laboratory experiments. They occurred with levels of legacy P that seem high. However, these results were obtained for soil with very high P buffering. Figure 1b shows that to obtain good growth for the soil used, an application of about 150 mg pot−1 was needed. The concentration in the 200 g of soil to which it was applied was therefore 750 mg kg−1. It would not take many years of reapplication, even allowing for removal of P in produce, to reach the levels used.

Fig. 1
figure 1

Effect of previously applied phosphorus (legacy P) on plots of sorption against concentration (a) and on the response to further applications of phosphate (b) To establish the levels of legacy P, the indicated amounts of P (mg kg−1) were incubated with moist soil at 70 °C for 30 days This is equivalent to about five years at 20 °C (Barrow et al. 2022) Sorption was then measured by gently mixing samples of the soil with a range of P concentrations in a 00 1 M CaCl2 solution Plant response was measured in a pot trial in which the indicated levels of P were added to 200 g samples of the soil and placed in the pot in a layer Sorption (S) was described by an equation of the form: \(S=a {x}^{b}-q\) where x indicates the concentration, and the other symbols indicate parameters, explained in the main text Plant response was described using Eq. (1): \(y=m (1-\text{exp}\left(-c \left(x+d\right)\right)\) (Redrawn from Barrow et al. (2022))

Full size image
Fig. 2
figure 2

Effect of previously applied phosphorus (legacy P) on the rate of dissolution of iron and aluminium from a soil. The levels of legacy P were established as for Fig. 1 Samples of the soil were then extracted with either 5 M HCl or 01 M NaOH The values plotted are for the cumulative amounts dissolved (Redrawn from Barrow et al. (2023)

Full size image
Fig. 3
figure 3

Effect of previously applied phosphorus (legacy P) on plots of sorption and desorption against concentration (parts a and b) and on plant response (parts c and d) The indicated levels of legacy P were established as for Fig. 1, and sorption was measured as in that figure The supernatant solution was then removed and replaced with a 001 M CaCl2 containing no P, to measure desorption For the pot trial, the indicated levels of P (mg pot−1) were added to 200 g of soil and incubated at 70 °C for the indicated periods Each day at this temperature is equivalent to about 60 days at 20 °C (Barrow et al. 2022) The treated soil was then added to the pots in a band Sorption (S) and the desorption were described by equations of the form: \({S}_{\text{s}}={{a}_{s} {c}_{s}}^{b}-{q}_{s}\) and \({S}_{\text{d}}={{a}_{d} {c}_{d}a}^{b}-{q}_{d}\) where c indicates the concentration, the other symbols indicate parameters and the subscript s and d indicate sorption and desorption phases. lant response was described using Eq. (1): \(y=m (1-\text{exp}\left(-c \left(x+d\right)\right)\) (Redrawn from Barrow et al. (2022))

Full size image
Fig. 4
figure 4

Summarising the effects of legacy P on plant growth and on sorption measurements In a, the plant property plotted is the reciprocal of the parameter c of Eq. 1 (Fig. 1) This is proportional to the P required for any given fraction of the maximum yield The sorption property is the parameter a of Eq. (2) (Fig. 1) In b, the plant property plotted is the relative rate of decline in effectiveness due to continuing reaction before the plants were grown(Details are given in Barrow et al. (2022) The sorption property the hysteresis ratio given by ad / as (Fig. 3) A value of unity indicates no hysteresis

Full size image

Soil tests tend to measure quantities. That is why they are frequently (and unfortunately) called measures of “available P”. They do not measure the important changes in P characteristics of soils following long-term phosphate application. They were calibrated when the soil phosphate stocks were much lower. Those calibrations may overestimate the current need for phosphate. This is a major reason for over-application of phosphate fertilisers.

We may not be able to readily access the legacy P, but we should recognise that it has large effects, decreasing the current need for phosphate.

It is frustrating that so much research is devoted to seeking a modern philosophers stone, but so little to understanding the nature of soil reaction with phosphate and the effects on the current need for phosphate fertilisers. The potential rewards are much bigger.