The Origin of Oxygenic Photosynthesis

"It is hard to imagine a more dramatic step in biochemical evolution than the final tinkering with the manganese–calcium cluster in the oxygen-evolving complex of PSII that led to the splitting of water and the release of molecular O2."
Joseph Kirschvink and Robert Kopp, Philosophical Transactions of the Royal Society B, 2008¹

"there are no bacterial lineages known that branch immediately ancestral to cyanobacteria and that possess only one RC or some other 'primitive' form of photosynthesis."
William Martin, Donald Bryant and J. Thomas Beatty, FEMS Microbiology Reviews, 2018²

Summary

Almost all the oxygen in the air was made by one machine.  Cyanobacteria, algae and plants use light to pull electrons out of water, and the oxygen is what is left over.  The work needs two light-driven protein complexes, Photosystem II and Photosystem I, working in series, and a cluster of four manganese atoms and one calcium atom that holds the water while four electrons are taken from it.  Every other photosynthetic bacterium uses a single, simpler reaction centre and makes no oxygen.  This article asks how the oxygen-making machine arose, and for each step separates what has been observed, what is inferred and what is assumed.

1. The machine is large, and it destroys itself in use. Photosystem II has about 20 protein subunits and 35 chlorophylls in each half of its paired form. Seven amino acids, six from the D1 protein and one from CP43, hold the manganese cluster.³ Photosystem I has 12 subunits and 127 cofactors.⁴ More than 20 helper proteins guide the assembly of Photosystem II.⁵ Light damages it at every intensity, so cells keep taking it apart and replacing its D1 protein.⁶
2. The creationist and design case is mostly an argument from the whole system. Rick Swindell, writing in Journal of Creation in 2003, argues that photosynthesis is irreducibly complex, that "nothing works unless everything works," and that "there has never been a time in Earth's history when there was not significant free oxygen."⁷⁷⁷ Answers in Genesis repeats him. Cornelius Hunter of the Discovery Institute says cyanobacteria "just appeared" with the machinery and that its source is unknown.⁸ Writers at Reasons to Believe and ISCAST accept the mainstream history and credit design. No young-earth or design writer found here discusses where the manganese cluster came from or the engineered reaction centres, and only Hunter mentions the cyanobacteria's non-photosynthetic relatives.
3. The two photosystems are related to simpler reaction centres. That much is observed. The core proteins of Photosystem II share about 17% of their amino acids with those of the purple bacteria's reaction centre, and the lead author of that count writes that "There is no doubt" they share a common origin. Photosystem I has the same build as the single photosystem of green sulfur bacteria.⁹¹⁰ Each simpler type runs photosynthesis by itself in living bacteria today. Photosystem I also runs alone in plants when Photosystem II is blocked, cycling electrons to make ATP.⁹ So Swindell's "nothing works unless everything works" is wrong for photosynthesis in general. It is right for the one job of taking electrons from water all the way to the cell's reducing molecule, which needs both photosystems.
4. Nobody has seen an intermediate. No living organism has a water-splitting centre in a simpler form, a single reaction centre that oxidises manganese, or the two photosystems under a switch. The closest living relatives of cyanobacteria, found in 2013 and 2017, have no photosynthetic genes at all.¹¹¹² Whether cyanobacteria got their two photosystems by merging genes from two other bacteria or by duplicating one and losing parts elsewhere is still argued, and Martin, Bryant and Beatty write that there is "no consensus."²
5. Laboratories have built the first step and stopped there. JoAnn Williams and James Allen gave a purple-bacterial reaction centre six amino-acid changes, three to make it more oxidising and three to make a manganese site. It binds one manganese, takes an electron from it, and turns superoxide into oxygen. It does not split water, and without the manganese the more oxidising version is not stable in light.¹³¹³ The real cluster assembles itself on its protein in light, with about 1% efficiency.¹⁴ No experiment has evolved a water-splitting centre from one that cannot split water.
6. The first step can be priced, and the price depends on whether each change helped. The six changes need at least 11 DNA base changes. At the one measured cyanobacterial mutation rate, if each change helped its owner a little, all of them spread in a few centuries. If each was neutral and drifted in alone, about 50 to 70 million years. If all had to arrive together, as in Michael Behe and David Snoke's model, the manganese site alone takes about 30,000 years in a population the size of today's ocean cyanobacteria and far longer than the age of the universe in a population of the size that sets the pace of drift. The whole machine has no counted pathway, so it cannot be priced at all.
7. The dates are not settled, and the rocks give an order. Mainstream estimates for when water splitting began run from 3.8 to about 2.2 billion years ago, and molecular clocks give dates from 3.5 to 1.5 billion years.¹¹⁵ The rocks show an air with almost no oxygen, then oxygen, and Stromatolites and Banded Iron Formations tests the young-earth placements of those rocks. Swindell's claim that oxygen was always present contradicts the textbook he cites for it.⁹

The verdict.  Three questions get three answers.  The claim that photosynthesis is irreducible because nothing works without everything is wrong:  simpler reaction centres work alone, the parts are related, and several outer subunits can be removed one at a time.  The claim that no stepwise route to the water-splitting machine has been shown is right:  no organism, experiment or calculation has gone past the first manganese, and the mainstream papers say so.  When the machine arose, in which lineage, and by merger or by duplication, is unfinished on both sides, and no young-earth model has yet tried to explain the evidence the mainstream argues from.

Part 1: What has to exist

Two photosystems in series.  Oxygenic photosynthesis takes electrons from water and puts them on NADP+, the carrier the cell uses to build sugar.  Robert Blankenship's standard textbook describes the main route as "non-cyclic electron transfer, in which water is oxidized to molecular oxygen and NADP+ is reduced to NADPH, driven by two sequential photoreactions in photosystems 2 and 1."⁹  Photosystem II makes an oxidant strong enough to pull electrons out of water.  Photosystem I makes a reductant strong enough to load them onto ferredoxin and NADP+.  Each electron is pushed uphill twice, once by each photosystem, so the minimum is eight photons for each oxygen molecule.⁹  This arrangement is called the Z scheme, after the shape of its energy diagram.

A very strong oxidant.  Water gives up its electrons only to an oxidant stronger than +0.82 volts at neutral pH.⁹  The oxidised chlorophyll pair at the heart of Photosystem II, P680, reaches an estimated +1.2 volts, which Blankenship states without a source.⁹  The equivalent pair in the purple bacteria's reaction centre sits at about +0.5 volts.¹⁶  So the purple-bacterial centre, the nearest relative of Photosystem II, falls about 0.7 volts short.

A four-electron store.  Water oxidation removes four electrons to make one oxygen molecule, but a reaction centre moves one electron per photon.  Blankenship calls this "a serious mismatch between the intrinsic chemistries of these two processes."⁹  The fix is a cluster of four manganese atoms and one calcium atom, bridged by five oxygen atoms, which steps through a cycle of states and gives up oxygen only after the fourth.³¹⁷  Six amino acids of the D1 protein and one of CP43 hold the metals, and each manganese has six attached atoms and the calcium seven.³³  Two chloride ions sit close by, and Ca2+ and Cl- are both needed for the cluster to assemble and work.¹⁸

The parts list.  Each half of the paired Photosystem II holds about 20 protein subunits, 35 chlorophylls, 2 pheophytins, 11 carotenes, 2 plastoquinones, a non-haem iron, two haems and the metal cluster.³  Photosystem I has 12 subunits and 127 cofactors, 96 of them chlorophylls.⁴  Light-harvesting antennas feed both.  In cyanobacteria the antenna is the phycobilisome, a protein assembly of 5 to 20 megadaltons on the outside of the membrane.¹⁹  In plants it is LHC-II, which binds 14 chlorophylls per unit.²⁰  The light reactions also need the cytochrome b6f complex, plastocyanin, ferredoxin and an ATP synthase, which this article does not count.

Assembly with helpers.  Photosystem II is not put together in one step.  More than 20 helper proteins that are not part of the finished machine guide its assembly in modules.⁵  One of them, Psb28, holds the half-built complex in a protected state in which the iron near the electron acceptor is bound by a glutamate, as it is in the purple bacteria's reaction centre.⁵  The metal cluster is added last.  The D1 protein is made with a tail that a protease must cut off before the cluster can form.¹⁸

The cluster builds itself, slowly.  When the cluster is stripped out of Photosystem II, the amino acids that held it stay where they were, leaving "a pre-organized ligand shell."²¹  Isolated Photosystem II stripped of its cluster rebuilds it in the test tube from manganese and calcium ions in light, a process called photoactivation.¹⁴  Bicarbonate and the outer protein subunits raise its speed and yield.¹⁸  A sub-complex lacking CP43, the protein that supplies one of the seven ligands, can still be photoactivated, but it makes oxygen at 20% of the normal rate and is "highly unstable."¹⁸  It is inefficient:  about 1% of absorbed photons lead to assembly, against more than 90% for water splitting once the cluster is built.¹⁴  Without calcium the protein binds and oxidises up to 20 manganese ions and makes no oxygen.¹⁸

A machine that wrecks itself.  The strong oxidant that splits water also damages the protein around it.  Light damage to Photosystem II happens at every light intensity, and the cell repairs it by partly taking the complex apart, replacing D1 with a new copy, and putting it back together.⁶  Gerard Dismukes's group puts the loss of activity "as often as every 15-30 minutes under normal solar flux conditions."¹⁸  In a strain of the cyanobacterium Synechocystis grown in dim light, D1 was replaced five to twenty times faster than the other Photosystem II proteins, with a half-life of one to one and a half hours.²²  Because D1 holds most of the cluster's ligands, the cluster has to be rebuilt each time.¹⁸

Which parts are needed.  Not every subunit is needed for growth.  Cyanobacteria with any one of the three outer proteins PsbO, PsbU or PsbV deleted still grow on light alone, more slowly:  doubling times rose from about 12 hours to 16 hours without PsbV and 19 hours without PsbO.²³  Remove PsbO and PsbV together and the cells cannot grow on light, with less than 10% of normal oxygen output.²⁴  The core proteins that hold the chlorophylls and the cluster have no such slack:  without them there is no Photosystem II.

What this part shows.  All of the above is observed.  It sets the bar any origin account has to clear:  a strong oxidant about 0.7 volts beyond the nearest relative's, a four-electron metal store held by seven side chains from two proteins, an assembly line, and a repair cycle fast enough to replace the core every hour or so.

Part 2: What creationists and design writers say

The young-earth literature on this topic is small.  Searches of Creation Ministries International, Answers in Genesis and its Answers Research Journal, the Institute for Creation Research, the Creation Research Society Quarterly, the Discovery Institute's sites, the journal BIO-Complexity and Reasons to Believe found the writers below.  Their claims, in their own words, fall into four groups.

1. The whole system had to arise at once.  Rick Swindell, a plant geneticist writing in Journal of Creation in 2003, gives the fullest case.  His thesis is "that evolution would not have been capable of generating the process of photosynthesis as it exists in cyanobacteria, green algae and higher plants, and that it must therefore have been intelligently designed."⁷  Most of his paper is about the 17 enzymes that make chlorophyll, the Calvin cycle and the enzymes that protect the cell from oxygen.  On the two photosystems he writes:  "Neither system does anything meaningful apart from the other, which is to say, nothing works unless everything works."⁷  He concludes that "It defies common sense to imagine that the irreducible complexity of photosynthetic systems would have arisen according to evolutionary theory."⁷  Roger Patterson's Evolution Exposed for Answers in Genesis summarises Swindell:  "The evolution of photosynthesis appears to be an impossible process as the intermediate products are toxic to the cell."²⁵  Answers in Genesis's children's page says that "God created the various plant kinds with the ability to photosynthesize from the beginning."²⁶  Frank Sherwin of the Institute for Creation Research writes that God made plants on Day 3 "with all of their structures (e.g. chloroplasts) and metabolic processes complete and fully functioning."²⁷

2. No mechanism has been given.  Georgia Purdom of Answers in Genesis, reviewing a 2007 Nature essay by John Allen and William Martin, writes:  "No mechanism is provided for how the photosystems came into existence, just that 'these genes arose' for one photosystem."²⁸  She objects that duplicating a photosystem's genes and mutating them "will only lead to the loss of information," and she asks a sharper question:  "what is the selection pressure to keep both photosystems in a bacterium if one is sufficient?"²⁸  Cornelius Hunter, a Discovery Institute fellow, made the same point in 2017 about the discovery that the closest relatives of cyanobacteria cannot photosynthesise:  "The problem is, from where did the photosynthesis come?"⁸  He calls convergence, gene transfer and fusion "made up mechanisms to fix the problem that the scientific evidence contradicts evolutionary theory."⁸

3. Oxygen was always there, and would have killed the first photosynthesisers.  Swindell writes that "there has never been a time in Earth's history when there was not significant free oxygen in the atmosphere," and that "the appearance of ground state oxygen would have been lethal to the cell well before oxygen-disarming complexes evolved."⁷⁷  Patterson draws the young-earth conclusion:  "If photosynthesis was present from the beginning, the 'oxygen revolution' was never a condition that evolving life had to deal with."²⁵  Michael Oard in Journal of Creation argues that the sulfur-isotope evidence for an oxygen-free early atmosphere "is now equivocal."²⁹  The only young-earth writer found who treats the Great Oxidation Event as an open question rather than an error is Johanna Wood, in a 2025 student poster abstract, who says that "the GOE's validity in earth history would also need to be addressed in a thorough Biblical model."³⁰

4. Accept the history and credit design.  Hugh Ross of the old-earth ministry Reasons to Believe accepts a Great Oxidation Event about 2.4 billion years ago and early photosynthesis, and writes that "Thanks to the many supernatural interventions on the part of the Creator, all these conditions were met and the Great Oxygenation Event did indeed occur."³¹  Trevor Rapson, a CSIRO chemist writing for the Christian science fellowship ISCAST, accepts the mainstream account and argues from the precision of the machine:  "First, PS II exhibits remarkable fine-tuning at a sub-nanometre scale:  each atom is precisely positioned.  Second, despite this exquisite design, PS II is irreversibly damaged in the very act of splitting water, requiring continual replacement."³²

What none of them discusses.  No young-earth or intelligent-design source found says anything about where the manganese cluster came from.  Swindell describes it in one sentence and moves on.  None discusses the engineered manganese reaction centres, the molecular-clock dating of the photosystems or the manganese-precursor hypothesis, and only Hunter's news commentary discusses the non-photosynthetic relatives of the cyanobacteria.  The Discovery Institute's design theorists have written at length on the bacterial flagellum and blood clotting, but no paper by Michael Behe, Douglas Axe, Ann Gauger or Stephen Meyer on photosynthesis was found.  Robert Blankenship wrote in 2008 that photosynthesis "has been portrayed in ID literature as irreducibly complex in those aspects that are not well understood," but he named no design writer, and none who says so was found.³³  The argument about irreducibility in general belongs to Irreducible Complexity, and this article tests only its application here.

Errors in the sources.  Three are plain enough to note before the evidence.  Sherwin twice puts oxygen release in the "light-independent stage (the so called dark phase)," with the sugars.³⁴²⁷  Oxygen comes from water splitting in Photosystem II, which is a light reaction, and Swindell's own paper says so.⁷  Sherwin's 2019 article says evolutionists suggest "this complex process may have spontaneously begun 3.4 billion years ago."³⁴  His source was a press release about Tanai Cardona's 2018 paper, which dates a gene duplication in Photosystem I, not the origin of photosynthesis, and the word "spontaneously" appears in neither.³⁵  Swindell's claim that oxygen was always present cites Blankenship's textbook, which says the opposite (Part 7).

The claims to test.  From these sources, four testable claims follow.  First, that the photosystems work only together and so could not have arisen in parts.  Second, that no stepwise mechanism for the water-splitting machine has been given or shown.  Third, that duplicating and changing genes cannot make a new photosystem.  Fourth, that oxygen was present from the start, so there was no oxygen-free world in which the machine could have evolved.  The first is tested in Part 3, the second in Parts 4 to 6, the third in Parts 3 and 6, and the fourth in Part 7.

Part 3: Related machines, and how two came to work together

Observed:  the reaction centres resemble each other.  Every photosynthetic bacterium has one of two kinds of reaction centre, called type I and type II.  Photosystem II is a type II centre, like the one in purple bacteria.  Photosystem I is a type I centre, like the one in green sulfur bacteria and heliobacteria.  Within the type II family the resemblance shows in the amino acids.  Tanai Cardona and colleagues report that "Any standard D1 or D2 shares on average 17% sequence identity with any L or M," where D1 and D2 are the core of Photosystem II and L and M the core of the purple bacteria's centre, and they write that "There is no doubt that D1, D2, L, and M share a common origin."¹⁰  JoAnn Williams and James Allen, who redesign the purple-bacterial centre, begin a paper:  "Although evolutionarily related to photosystem II, bacterial reaction centers lack both a strong oxidant and a manganese cluster."³⁶  Between the two types the resemblance is in shape only.  Martin, Bryant and Beatty write that they "are related at the level of structure ... but with undetectable homology in amino acid sequence alignments."²

Inferred:  common descent of the reaction centres.  Sequence identity and shape are measurements.  That the centres descend from one ancestral protein is an inference from them, and within the type II family it is about as strong as such inferences get.  A design account can accept the resemblance and credit a common designer.  It then owes an answer to the details that do no work.  One is the half-built Photosystem II held by the assembly protein Psb28, in which the iron near the acceptor end is bound by a glutamate, "a structural motif found in reaction centres of non-oxygenic photosynthetic bacteria."⁵  The same pattern was found for organelles in Endosymbiosis and the Eukaryotic Cell.  The link between type I and type II is much weaker, because it rests on structure alone.

Observed:  each kind works alone.  Purple bacteria run photosynthesis on one type II centre, cycling electrons to make ATP.²  Green sulfur bacteria "have a single photosystem similar to photosystem 1" that loads electrons onto ferredoxin directly.⁹  Even in plants, Photosystem I runs alone when Photosystem II is blocked with a herbicide, cycling electrons to make ATP.⁹  So the claim that "Neither system does anything meaningful apart from the other" is wrong as a general statement.⁷  What is true is narrower:  taking electrons from water and delivering them to NADP+ needs both, because one photon cannot lift an electron that far.  That is the step that has to be explained.

Not observed:  an organism partway.  John Allen proposed in 2005 that both photosystems first existed in one anaerobic lineage, switched on and off by the cell's redox state, "living examples of which are predicted," and that oxygenic photosynthesis began with "a simple mutation that disabled the redox switch."³⁷  No such organism has been found.  Martin, Bryant and Beatty, who favour a similar scenario, write that "there are no bacterial lineages known that branch immediately ancestral to cyanobacteria and that possess only one RC or some other 'primitive' form of photosynthesis."²

Inferred, and disputed:  merger or duplication.  Filipa Sousa, John Allen, William Martin and a colleague set out the two main accounts in 2013.  Either the two kinds of centre evolved in different bacteria "and became merged in cyanobacteria via an event of large-scale lateral gene transfer," or "the two RC types are related via gene duplication in an ancestral, anoxygenic but protocyanobacterial phototroph," after which other bacteria lost one kind or the other.³⁸  Their evidence for duplication is that the genes for making chlorophyll show "no dichotomy" into two families, one for each kind of centre, which a merger of two lineages would predict.³⁸  Armen Mulkidjanian and colleagues proposed a third version, in which photosynthesis began in the cyanobacterial line and spread outward.  They warn that photosynthesis genes move between distant bacteria so readily that this "makes identification of the lineage that was the first to develop chlorophyll-based photosynthesis particularly challenging."³⁹  Gene transfer of photosystem genes is itself observed:  viruses of ocean cyanobacteria carry the Photosystem II core genes and swap them with their hosts "from host to phage, phage to host, and within the phage gene pool."⁴⁰  But those are genes moving between cyanobacteria that already have Photosystem II.  Nobody has seen a photosystem move into a lineage that lacked one.

Martin, Bryant and Beatty sum up the state of the question:  "There are two main questions about the origin of two types of RCs:  how and why?  The former is much easier to answer than the latter."²  Purdom's question about why a cell would keep two photosystems is the same question.  Martin's group answers it in principle:  two kinds of centre "could be differentially expressed to serve ATP synthesis or Fd reduction, as needed by the cell."²  That is a reason that would work if the organism existed.

Observed:  the closest relatives have nothing.  In 2013 Sara Di Rienzi and colleagues rebuilt the genomes of bacteria from human guts and groundwater that branch next to the cyanobacteria, the Melainabacteria, and found no photosynthetic genes.  They inferred that oxygenic photosynthesis did not arise before the two lines split, "based on the absence of remnants of photosynthetic machinery."¹¹  In 2017 Rochelle Soo and colleagues added a second, deeper group, the Sericytochromatia, and concluded that the cyanobacteria "acquired the genes for photosynthesis relatively late in cyanobacterial evolution."¹²  This is the finding Hunter took as showing the machinery has no source.  The Imperial College group draws the opposite conclusion from the same tree:  "Vampirovibrionia, Sericytochromatia and Margulisbacteria lost photosynthesis."⁴¹  Martin's group calls it "presently unsubstantiated" to name these lineages cyanobacteria at all.²  The genomes are observed.  Whether the relatives never had photosynthesis or lost it is inferred, and the two inferences contradict each other.  On either reading, no living relative shows a halfway machine.

What this part shows.  The parts are related, and each kind of reaction centre works on its own in some living bacterium, so the strong form of the irreducibility claim fails.  The weaker form survives:  no organism has been found that holds an intermediate between one centre and two in series, and the leading mainstream reviewers say the "why" of the joining is the harder question.

Part 4: From one manganese to water splitting

The hardest step is the one Kirschvink and Kopp call "the final tinkering with the manganese–calcium cluster."¹  The mainstream literature says plainly that it is not understood.  Cardona, James Murray and William Rutherford:  "it is still unclear how an ancestral bacterium evolved the capacity to split water."⁴²  Jena Johnson, Woodward Fischer and colleagues:  "the processes that led to the evolution of biological water splitting have remained largely unknown."⁴³  Cardona and colleagues in 2019:  "how and when one of these reaction centers evolved the capacity to oxidize water are questions that still remain to be answered."¹⁰

The proposals.  Several precursors have been proposed, each a simpler electron donor that a reaction centre could have oxidised before water.  Robert Blankenship and Hyman Hartman proposed hydrogen peroxide.  Their summary of what the change needs is short:  "Two major changes from a more primitive non-oxygen-evolving reaction center are required:  a charge-accumulating system and a reaction center pigment with a greater oxidizing potential."⁴⁴  Charles Dismukes's group proposed bicarbonate and manganese-bicarbonate complexes.⁴⁵  The proposal with the most work behind it is a reaction centre that oxidised dissolved manganese one electron at a time.  Cardona's group writes that "The earliest reaction center probably performed the light-driven one-electron oxidation of Mn2+ to Mn3+."⁴²  On this view the manganese site came first as an electron source, and the cluster later learned to hold four charges and make oxygen.

The rock evidence for a manganese stage is disputed.  Johnson and colleagues found manganese enriched to about 17% by weight in South African drill cores just below the Great Oxidation Event, and argued that the original manganese oxides "were not produced by reactions with O2, which points to a different high-potential oxidant," supporting "a former transitional photosystem capable of single-electron oxidation reactions of Mn."⁴³  CarriAyne Jones and Sean Crowe replied that oxygen was not ruled out, so "the conclusion that Mn-oxidizing photosynthesis produced the Koegas Mn deposits is highly tenuous."⁴⁶  Then in 2019 Mirna Daye, Tanja Bosak and colleagues showed that a lake community of anoxygenic bacteria makes manganese oxides in light with no high-potential reaction centre at all.  Before their work, "modern microbes are not known to anaerobically oxidize manganese."⁴⁷  Their green sulfur bacterium's reaction centre, at about +250 millivolts, "cannot directly oxidize Mn(II)-bicarbonate," and the activity needed a second species, by "a currently unknown mechanism."⁴⁷  So manganese oxides in old rocks need not mean a manganese-oxidising photosystem, and no living organism has one.

Laboratories have built the first step.  Over nearly two decades the Arizona State group of JoAnn Williams and James Allen rebuilt the purple-bacterial reaction centre toward Photosystem II.

• Adding hydrogen bonds to the chlorophyll pair raised its potential. The effects of single changes added up, giving potentials from 410 to 765 millivolts against 505 for the wild type, and the bacteria still grew by photosynthesis, "although at reduced rates relative to the wild type."¹⁶
• Adding carboxylate side chains where Photosystem II holds its cluster made a site that binds one manganese ion. In the high-potential centres the bound manganese gives up an electron to the chlorophyll pair within milliseconds.³⁶ The measured potential of that manganese, 625 millivolts at pH 9, sits below the potential the mutant centres reach, which runs to above 800 millivolts.⁴⁸
• In 2012 they reported that these centres, with six changes in the M8 version, turn superoxide into oxygen in light, "a light-driven enzymatic function."¹³¹³ The more oxidising centres "are not stable in the light unless they have a bound Mn."¹³

That is a real result.  It shows that a reaction centre from a bacterium that makes no oxygen can be given a manganese site and a stronger oxidant with a handful of changes, and that the manganese then does useful chemistry.  The authors draw the evolutionary lesson in the conditional:  such substitutions "would have provided a metal-binding site and consequently resulted in the production of molecular oxygen from any superoxide that was present at that time."¹³  Three limits matter.  The changes were chosen by people who knew the target.  The centre holds one manganese and moves one electron, not four.  And it does not oxidise water:  its strongest form, above 800 millivolts, is still about 400 millivolts short of P680.

A second line of engineering put a pair of manganese ions and a light-absorbing pigment into bacterioferritin, an iron-storage protein from E. coli, and light then oxidised the manganese pair.⁴⁹  Its own designers' calculation puts the pigment's potential at "only 0.57 V, too low to oxidize tyrosine," and says that to "ultimately oxidize water" it would need "a value in excess of 0.8 V."⁵⁰

Today's Photosystem II can make manganese oxide.  In 2020 Holger Dau's group stripped the cluster from spinach Photosystem II and gave it dissolved manganese in light.  It made particles of manganese oxide of 50 to 100 manganese ions each, bound to the protein.⁵¹  They propose that an ancestral photosystem made such oxides and that the particles were later "down-sized" to the four-manganese cluster.  They also write that "The evolutionary route towards the present water oxidation catalyst in PSII is much debated."⁵¹  The experiment used a modern photosystem, which already has P680's oxidising power.  It shows the chemistry such an ancestor would need is possible in that protein.  It does not show the ancestor.

What has not been done.  No laboratory has started from a reaction centre that cannot split water and evolved or engineered one that can.  No organism alive oxidises manganese with a photosystem.  The step from one manganese and one electron to four manganese, a calcium and four stored charges is described in the literature as possible, not shown.  Cardona's group:  "it is possible that early transitional forms were capable of some inefficient water oxidation, even with a mononuclear manganese cluster."⁴²  Their 2023 review:  "It still remains unclear whether Mn2+ oxidation without water oxidation occurred as a transitional stage in the evolution of PSII."¹⁵

What this part shows.  The first step toward the cluster, a manganese site in a reaction centre with a stronger oxidant, has been built in the laboratory and does useful chemistry.  Every step after it is inferred from the structure of the finished machine.  On this point the creationist claim that no stepwise mechanism has been shown is correct, and the mainstream papers say the same thing in their own words.

Part 5: When, by the mainstream's own clocks

All dates in this part are the mainstream radiometric and molecular-clock figures under discussion.  Young-earth readers reject the time scale, and every position below is billions of years from a young earth.  The point here is how far the mainstream positions are from each other, and what each one assumes.

The range is wide.  Tanai Cardona wrote in 2018 that "When and how oxygenic photosynthesis originated remains a highly debated subject with dates ranging from 3.8 billion years (Ga) ... to shortly before 2.4 Ga."³⁵  Kirschvink and Kopp give the same span, 3.8 to 2.2 billion years.¹  The Imperial College review of 2023 says molecular clocks give "a range of dates spanning two billion years between 3.5 and 1.5 Ga."¹⁵

The late view.  Kirschvink and Kopp argue that oxygenic photosynthesis arose shortly before a global glaciation about 2.3 billion years ago and helped cause it.¹  Patrick Shih, Woodward Fischer and colleagues dated the last common ancestor of today's photosynthetic cyanobacteria to "ca. 2.0 billion years ago (Ga), well after the rise of atmospheric dioxygen."⁵²  Johnson's group, replying to critics, notes the catch in using such dates:  "the divergence time between cyanobacteria and other phyla is not synonymous with the origin of oxygenic photosynthesis"⁵³

The early view.  Cardona's group dates the split of Photosystem II's two core proteins, D1 and D2, by how fast they have changed since.  They find that if reaction centres existed by 3.5 billion years ago, "the initial rate of evolution had to be about 40 times larger than that observed since the MRCA of Cyanobacteria."¹⁰  They grant the problem this raises:  "The much faster rates required to place the early stages of reaction center evolution ... in the Paleoarchean are difficult to reconcile with the structural complexity inherited by all known reaction centers."¹⁰  And they ask the reader to "interpret the reported age estimates as an approximation, as simulations of plausibility."¹⁰  The Photosystem I study assumes "that photochemical reaction centres originated in the early Archean" as an input to the clock.³⁵  A 2021 follow-up says "The origin of oxygenic photosynthesis cannot be located on a species tree," and that "different datasets, calibration strategies, chosen clock parameters, and researchers' preferences can result in widely varying calculated rates."⁴¹⁴¹

The newest large clock.  Adrián Davín, Gergely Szöllősi and 18 colleagues dated the bacterial tree in 2025 and found that "aerobic metabolism likely pre-dated the GOE in the cyanobacterial ancestor, which may have facilitated the evolution of oxygenic photosynthesis."⁵⁴  They add that "we cannot reject the possibility that oxygenic photosynthesis predated, or co-evolved with, aerobic metabolism in the cyanobacterial lineage."⁵⁴  Several of its authors also wrote the 2017 Science paper on the non-photosynthetic relatives, so the two are not independent.

A rate that prices time.  Cardona's group also asked how fast D1 and D2 would have had to change if water splitting arose quickly after a reaction centre arrived by gene transfer.  If it took 100,000 years, the rate "would need to be more than 400 thousand amino acid changes per site per Ga, which is orders of magnitude greater than the rate of evolution of any known protein."¹⁰  If it took 100 million years, about 33 changes per site per billion years.  That is the mainstream's own reason for wanting a long time between the first reaction centre and water splitting.  It also shows why a fast evolutionary origin is not on offer from any side.  A young-earth account does not need one, since it holds that the photosystems were made, not evolved.

What this part shows.  The order of events is argued from gene trees and molecular clocks, and the clocks disagree by more than a billion years depending on calibration, rate model and which lineage is taken to be the first oxygen maker.  None of them observes the origin.  They are estimates of when, under the assumption that it happened by descent, and their own authors describe them as approximate.

Part 6: Pricing the first step

Only one route toward Photosystem II has its steps written down as specific changes:  the Arizona State group's M8 reaction centre, with three changes to strengthen the oxidant and three to build a manganese site.¹³  It is a laboratory design, not a proposed history, and nothing says nature took it.  But it is the one stated pathway with a count, so it is the one that can be priced against measured mutation rates and population sizes.

The count.  Six amino-acid changes need more than six DNA changes, because most of them cannot be made by changing one letter of the codon.  In the published gene sequences of the purple bacterium Rhodobacter sphaeroides, the three potential-raising changes each need at least two base changes, two of the three manganese-site changes need two, and the last needs one.  So the manganese site needs at least 5 base changes, the stronger oxidant 6, and M8 as a whole 11.⁵⁵

The inputs.  The one measured mutation rate for a cyanobacterium is 3.5 × 10⁻¹⁰ per base per generation, from a three-year experiment on Prochlorococcus.⁵⁶  The same study puts its effective population size, the number that sets how strongly chance drives the fate of a neutral change, at 1.68 × 10⁷.⁵⁶  The census is far larger:  about 2.9 × 10²⁷ Prochlorococcus cells and 7 × 10²⁶ Synechococcus cells in today's oceans.⁵⁷  Nobody knows the population size of a bacterium before oxygenic photosynthesis, so the calculation runs across that whole range.  It assumes one generation a day.

Three answers, depending on the steps.  How long the changes take depends almost entirely on whether each one helped its owner.

1. If every change helped, by 1%, each spreads in turn. The 5 site changes take about 80 to 400 years, and all 11 about 170 to 870 years, across the range of population sizes.
2. If every change was neutral and drifted to fixation one at a time, the time does not depend on population size, because a neutral change fixes at the rate it arises. The 5 site changes take about 54 million years and all 11 about 71 million years. This is a floor, since it ignores changes drifting back.
3. If all the changes were needed together, with every partial state useless and at risk of being wrecked by other mutations, the right model is Michael Behe and David Snoke's. Their equations, recomputed here, reproduce their own printed figures to within about a factor of ten, erring low.⁵⁸⁵⁹ At the cyanobacterial mutation rate, the 5-base site takes about 29,000 years in a population the size of today's ocean cyanobacteria, about a million years in a population of 10²⁰, and about 10¹⁷ years, millions of times the age of the universe, in a population the size of Prochlorococcus's effective population. All 11 changes together take 10¹⁷ years even at the full census.

Which answer applies?  The laboratory results say something about this.  The changes that raise the potential are not neutral on their own:  the mutant bacteria grow more slowly, and the most oxidising centres "are not stable in the light unless they have a bound Mn."¹⁶¹³  So if this route was taken, the manganese site came first, and the potential-raising changes paid only if a bound manganese gave the cell some benefit.  The effect of the site changes alone on a living bacterium was not found reported.  The most favourable reading consistent with the data is that the 5 site changes drifted in as neutral changes over some tens of millions of years, and the 6 potential changes were then each favoured.  That fits easily inside the mainstream time scale.  The least favourable reading, all 11 needed at once, does not fit in any time scale.  Which reading is true has not been measured.

The limits of the price.  Behe and Snoke themselves warn that a result for one pathway says "little or nothing about the efficiency of other possible pathways," and that asking for one particular feature takes "a prospective stance ... but we look at modern proteins retrospectively."⁵⁸  Many different sets of changes might make a manganese site, and the Arizona group found two different mutants that work.¹³  Michael Lynch, replying to Behe and Snoke, argued that with neutral intermediate steps such features "can go to completion with high probability on time scales of 10⁶ yr or less in populations >10⁶ in size."⁶⁰  The wider argument over waiting times is in Haldane's Dilemma.

What cannot be priced.  Everything after the first manganese.  No paper proposes a sequence of protein states from a one-manganese centre to the four-manganese, one-calcium cluster with its two chlorides, the tyrosine relay, the outer proteins and an oxidant 400 millivolts stronger again.  Without a count of changes and a statement of which intermediates worked, there is nothing to price.  The creationist side has not proposed a count either.  This is the main gap in the evidence on both sides, and it is where a calculation would help most.

Part 7: What the rocks show, as order

Dates below are the mainstream radiometric ones, which young-earth geologists reject.  The order of the layers does not depend on the dates.  Whether the Precambrian rocks belong to Creation Week, the pre-Flood world or the Flood is tested in Stromatolites and Banded Iron Formations and The Earliest Evidence for Life, and is not argued again here.

Bottom:  air with almost no oxygen.  Sulfur in the older rocks carries an isotope pattern that forms only in sunlight acting on an atmosphere with almost no oxygen and no ozone.  Timothy Lyons and colleagues put the requirement at "probably less than 0.001% of the present atmospheric level."⁶¹  The pattern stops at the level called the Great Oxidation Event, about 2.4 billion years on the mainstream scale.

When oxygen production began is disputed.  Lyons and colleagues write that "Far more controversial is the timing of the first emergence of O2-producing photosynthesis, the source of essentially all oxygen in the atmosphere."⁶¹  The oldest chemical traces once credited to cyanobacteria did not survive testing.  A multilaboratory study with clean new cores concluded that "existing lipid biomarker evidence cannot be invoked to support the emergence of oxygenic photosynthesis and eukaryotes by ∼ 2.7 billion years ago."⁶²  Birger Rasmussen and colleagues put the oldest fossil evidence for cyanobacteria at "approximately 2.15 Gyr ago," above the oxygen rise.⁶³  The manganese evidence for earlier oxygen production, discussed in Part 4, is contested.

The young-earth claims about oxygen.  Swindell's statement that "there has never been a time in Earth's history when there was not significant free oxygen in the atmosphere" cites, among others, Blankenship's 2002 textbook.⁷  Blankenship says the opposite:  "Geological evidence strongly suggests that the atmosphere contained significant quantities of O2 by about 2.2 billion years ago," which "represents the latest date for the advent of oxygenic photosynthesis."⁹  Swindell's other cited source, Dimroth and Kimberley in 1976, argued from sediments before the sulfur-isotope evidence existed.  Oard's later reply to that evidence, that a laboratory reaction can make a similar sulfur pattern, is weighed in Stromatolites and Banded Iron Formations, which finds that it explains neither the ratio of the isotopes nor why the signal stops and never returns.

Swindell does point to a real puzzle.  Enzymes that use or detoxify oxygen are spread across the tree of life, which suggests they are older than oxygen-making photosynthesis.  Blankenship calls it "Another apparent paradox" and gives two answers:  "low levels of O2 and other reactive oxygen species were almost certainly produced on the early Earth by nonbiological processes, so even the earliest cells may have needed protection," or the protective enzymes arose after oxygenic photosynthesis and spread by gene transfer.⁹  Davín's 2025 clock also infers that the cyanobacterial ancestor used oxygen before the Great Oxidation Event.⁵⁴  So the claim that the first oxygen would have killed the first oxygen makers is answered in principle by low background oxygen.  Neither side has priced how much protection the first oxygen makers needed, or how much oxygen was around.

What the order shows.  It shows an early atmosphere with almost no oxygen, then oxygen, with the oldest recognisable cyanobacterial fossils at or above the change.  That is the order the mainstream account predicts.  A young-earth account has to place that oxygen-free atmosphere somewhere in its history, and Patterson's statement that the "oxygen revolution" was never a condition life faced does not do so.²⁵  The rocks cannot show how the machine arose.  They record its product, oxygen, and nothing about the protein that made it.

Part 8: Observed, inferred, assumed

The table shows where the evidence thins.  Everything about the finished machine is observed.  The kinship of its core to simpler reaction centres is a strong inference.  The path from a simple centre to a water-splitting one is assumed, apart from a first step built in the laboratory by design.

Objections and responses

1.  "Shared structure between Photosystem II and the purple-bacterial centre is common design, not common descent."  A designer could reuse a good fold, and a functional similarity fits both models.  The case for descent within the type II family rests on the amount of sequence shared, about 17%, together with features that do no job in the finished machine, such as the bacterial-type iron ligand that appears only in the half-built complex.  A design model can accept these after the fact.  It did not predict them, and no design writer found here has addressed them.

2.  "Swindell meant oxygenic photosynthesis, not photosynthesis in general.  For that, nothing works unless everything works."  That is the fairest reading, and in that form part of it holds:  taking electrons from water to NADP+ needs both photosystems.  But the claim was offered as a reason the system could not have been built in stages, and the stages proposed are each a working system:  one reaction centre alone, as in living bacteria, then two.  Even within Photosystem II, several outer proteins can be removed one at a time with growth continuing, and the cluster assembles itself on the protein.²¹²³  What remains without a demonstrated stage is the cluster and the oxidant, which is a narrower claim than "nothing works."

3.  "The engineered reaction centres were designed.  They show intelligence, not evolution."  True of the changes, which were chosen.  The experiment tests whether a manganese site and a stronger oxidant can be added to this protein with few changes, and whether the result does anything useful.  Both answers were yes.  It does not show that blind mutation found those changes, and Part 6 shows that the answer to that depends on whether each change helped, which has not been measured.

4.  "Duplication followed by mutation can only lose information."  Purdom states this without a measure of information.  D1 and D2 are the test case if their common origin is granted:  they share about 29% of their sequence and do different jobs, since D1 holds the manganese cluster and D2 does not.³¹⁰  Whether mutation made that difference cannot be observed now.  But "only lead to the loss of information" is a claim about what mutation can do, and it needs a definition of information and a calculation, neither of which Purdom gives.  The general question belongs to other articles in this series.

5.  "The closest relatives of cyanobacteria have no photosynthesis, so the machine just appeared."  Their genomes show that the machine is absent, not where it came from.  Soo's group reads the absence as a late gain by gene transfer, and the Imperial group reads it as loss.¹²⁴¹  On either reading, a gene-transfer donor or an ancestor with the machine is required, and neither has been found.  Hunter is right that the source is unknown.  He is wrong that gene transfer and fusion are "made up":  transfer of Photosystem II genes is observed today between cyanobacteria and their viruses.⁴⁰

6.  "Water splitting could have evolved through a manganese-oxidising stage, which the rocks record."  Possibly.  The rock evidence is disputed by other geochemists, and a 2019 culture study showed that anoxygenic bacteria can make manganese oxides by another route.⁴⁶⁴⁷  No organism with a manganese-oxidising photosystem is known.  The proposal is a hypothesis with a laboratory proof of concept for its first step, not a demonstrated pathway.

7.  "Deep time makes any number of steps possible."  Not without selection.  If the 11 base changes of the engineered route each helped, they take centuries.  If each was neutral, tens of millions of years.  If none helped until all were present, Behe and Snoke's own model puts the time for the 5-change site beyond the age of the universe in any population smaller than about 10¹⁴, and beyond 100 million years in any smaller than about 10¹⁶.  For all 11 changes the threshold is about 10³⁴, above any census of living cyanobacteria.  So time alone settles nothing.  What matters is whether the intermediate steps were useful, and the only data on that, from the high-potential mutants, show some of them were harmful on their own.

8.  "The pricing uses an arbitrary pathway."  It uses the only pathway written down as specific changes, which is also the one closest to working.  Behe and Snoke warn that a price for one pathway says little about others, and Allen's group found two different working mutants.⁵⁸¹³  Other routes may need fewer changes.  But nobody has written them down, and the creationist side has not proposed a count either.

9.  "Photosystem II wrecks itself every half hour.  A designer would not build it that way."  Rapson takes this feature as part of his design argument, and some critics take it as poor design.  The damage comes from the chemistry:  an oxidant strong enough to pull electrons from water will also oxidise the protein around it.  Whether a better design was possible is a question about chemistry neither side has answered.  The point matters here only because the repair cycle is part of what any origin account must produce.

10.  "The Great Oxidation Event is mainstream geology, so the young-earth side need not address it."  The young-earth side has to explain the rocks it accepts.  Rocks with the oxygen-free sulfur signal lie below rocks without it everywhere they have been studied.  Wood's 2025 abstract is right that a biblical model would need to address this.  None found here has.

Issues and unanswered questions

1. No stepwise pathway from one manganese to the water-splitting cluster has been written down. A proposal would list the protein states in order, say which ones work, and give the changes between them. With that, the waiting times could be priced as Part 6 prices the first step. Neither the mainstream nor the creationist literature has one.
2. The fitness of the intermediate steps is unmeasured. The pricing turns on whether a bacterium with only the manganese-site changes, or with one or two of the potential-raising changes, does better or worse than the wild type in light. The engineered strains exist. Competition experiments in light, with and without manganese, would answer it.
3. The residue number of one engineered change is uncertain. Allen and colleagues print "Gly M188 to Asp," but residue M188 in the published sequence is an asparagine and M288 a glycine. The count of base changes is the same either way. Thielges 2005 and Kálmán 2011, which may give the residue, were read only in abstract.
4. No living organism is known with an intermediate photosystem, whether a manganese-oxidising one, a homodimeric water splitter, or two photosystems under a switch. The mainstream accounts predict one may have existed. Finding one would settle much of this article.
5. Merger or duplication is unresolved, and the reading of the non-photosynthetic relatives is contradictory: Soo's group says gained late, the Imperial group says lost.
6. The dates rest on clocks whose authors call them approximate, and they span more than a billion years. The young-earth side has not engaged the clock papers at all.
7. The young-earth literature has not addressed the molecular evidence. No creationist or design source found discusses the manganese cluster's origin, the engineered reaction centres, the Melainabacteria and Sericytochromatia, or the clocks. The newest young-earth treatment of the oxygen record found is a 2025 student abstract that asks the question without answering it.
8. Errors in creationist sources need correcting at the source. ICR places oxygen release in the dark reactions, in Sherwin's articles of 2019 and 2023. ICR's "spontaneously begun 3.4 billion years ago" turns a Photosystem I duplication date into a claim about the origin of photosynthesis that no source makes. Swindell's "never a time" without significant oxygen contradicts the textbook he cites for it. Old-earth writer Hugh Ross's 2005 oxygen figures appear to come from a sulfate measurement in Kah and colleagues' 2004 paper, whose full text was not read.⁶⁴⁶⁵
9. Some numbers rest on abstracts or reviews. The Photosystem I counts, the Soo 2017 and Shih 2017 claims, and the engineered-centre potentials come from abstracts. The +1.2 volt figure for P680 is a textbook estimate with no reference. The D1 half-life is from a mutant strain in dim light; the faster figure for wild cells in brighter light was reported as data not shown.
10. The protection problem is not priced. How much oxygen-detoxifying capacity the first oxygen makers needed, and how much background oxygen there was, is argued in words on both sides.

Verdict

Is oxygenic photosynthesis irreducible, so that nothing works unless everything works?  The creationist argument is wrong as stated.  The two photosystems belong to families whose simpler members run photosynthesis alone in living bacteria, Photosystem I runs alone in plants when Photosystem II is blocked, several outer proteins of Photosystem II can be removed one at a time, and the manganese cluster assembles itself on its protein in light.  The kinship of Photosystem II's core with the purple bacteria's reaction centre is about as well supported as such inferences get.  What is true in Swindell's claim is narrower:  taking electrons all the way from water to NADP+ needs both photosystems and the cluster.

Has a stepwise route to the water-splitting machine been shown?  The creationist argument is right that it has not.  No living organism has an intermediate.  The laboratory has built the first step, a manganese site and a stronger oxidant added to a bacterial reaction centre by six designed changes, and it does useful chemistry, but it oxidises one manganese and not water.  Everything between that and the four-manganese cluster is inferred from the finished machine.  The mainstream papers say this themselves:  "it is still unclear how an ancestral bacterium evolved the capacity to split water."⁴²  The first step can be priced, and the price ranges from centuries to longer than the age of the universe depending on whether each change helped, which has not been measured.  The rest cannot be priced because no one has written down its steps.

When, in which lineage, and by merger or duplication?  Unfinished on both sides.  The clocks disagree by more than a billion years, the reading of the cyanobacteria's non-photosynthetic relatives is contradictory, and the manganese evidence in old rocks is disputed.  The rocks give a clear order, an oxygen-free atmosphere followed by an oxygenated one, which the young-earth side has not yet placed in its history.  The young-earth claim that oxygen was always present contradicts the source cited for it.

Sources