Two roads from N=8 SUGRA to string theory

Back in the early 1980s, people became excited about the N=8 supergravity theory (or "SUGRA" for short), a four-dimensional field theory that combines Einstein's general relativity with as much supersymmetry as you can get. Its existence looked somewhat non-trivial.

It was and it still is a beautiful structure. The high degree of supersymmetry made it more likely for divergences to cancel, a reason why some people use the term "the simplest quantum field theory" for this most supersymmetric field theory in four dimensions.

Soon it became clear that there were too major remaining problems with this supergravity, namely

1. the divergences cancel but not all of them
2. the structure was too constrained; N=8 supersymmetry surely doesn't allow you to have chiral fermions and similar things known from the Standard Model

Superstring theory has miraculously solved both problems which is why the high-energy theoretical physics community switched to the stringy mode of reasoning in the mid 1980s. During the last two decades, a huge amount of insights about the role of supergravity in the scheme of things has been discovered.

Concerning the first problem, people thought that the two-loop divergences would cancel but surviving divergences could occur at three loops. See

Finiteness of supergravity theories

for more details about the quantum loops where various gravitational divergences occur. This oldest guess was recently proven too pessimistic. As the hardcore technical papers by Bern, Dixon, Roiban, and others show, the three-loop divergences cancel, too.

Now it is likely that the first divergences occur at nine-loop level - see Green, Russo, Vanhove (2006) - or, which I find most likely, the theory is finite to all orders of perturbation theory.

See also: Lance Dixon's SUGRA puzzle

I find the "perturbatively finite" answer to be the most likely one. The most convincing general reason of the finiteness is that the N=8 gravity comes from a closed superstring much like the N=4 gauge theory arises from an open (or heterotic) superstring. The closed superstring includes left-moving and right-moving oscillations. They and their amplitudes are therefore related to second powers of the open (or heterotic) string amplitudes. Because the N=4 gauge theory is perturbatively finite, the N=8 supergravity may be perturbatively finite, too.

This relationship may be clarified in terms of

the Kawai-Lewellen-Tye (KLT) relations.

However, you should realize that the argument only works perturbatively because it is only the perturbative expansion where you can actually see the strings and argue that closed strings are open strings squared. These KLT relations tell you nothing about the non-perturbative behavior. As we will see later in the text, N=8 SUGRA is non-perturbatively inconsistent.




Besides the KLT relations, there are many other stringy methods (and corresponding papers) that are very helpful to determine the character of divergences in the N=8 supergravity, including a careful counting of the powers of the stringy parameter alpha' in various counterterms. See e.g. Bohr Jr & Vanhove (2008). One of the direct conclusions of the (surprising but observed) three-loop finiteness is that the cancellations can't be fully explained by supersymmetry.

The Bohr-Vanhove paper and others claim that the "higher" cancellations are partly due to general covariance and stringy constraints although the complete picture is not yet known. At any rate, it would be bizarre to use the new cancellations (surprising from a pure SUSY-QFT viewpoint) as an argument against string theory (that is probably the real reason why they take place).

Nevertheless, the problems 1,2 listed above haven't disappeared. The theory (pure supergravity) still breaks at the Planck scale - the perturbative series diverge, as explained below, and the theory cannot describe things like black hole production properly - and its phenomenology is not viable. I would like to explain that any intelligent enough person who accepts that the N=8 supergravity is on the right track inevitably ends up with string theory as her candidate unifying theory.

All roads lead to string theory

Such a person may decide to look for a fully consistent theory inspired by the N=8 supergravity or she may search for a phenomenologically acceptable generalization of the N=8 supergravity. In both cases, she is led to string theory. In order to see why, it is necessary to look at the two bugs of supergravity somewhat more accurately.

Non-perturbative completion

At the very beginning of this section, I (and a helpful anonymous commenter) feel it is important to emphasize that if a field theory is perturbatively finite - i.e. finite at every fixed order of perturbation theory - it doesn't mean that you can actually resum the series and obtain a fully finite result.

Quite on the contrary: in quantum field theory, and N=8 SUGRA is hardly an exception, the Taylor sums are almost always divergent (and in perturbative string theory, they are divergent, too). They are "asymptotic series". This divergence means that the perturbative result can't be the whole story but it doesn't mean that there can't exist a complete, finite answer.

Just on the contrary: a complete finite result does exist and the perturbative expansion is its Taylor approximation.

The minimum error you encounter by summing an "optimal" number of the perturbative terms to get a pretty accurate but convergent result (for a very small coupling constant) is of the same order as the leading non-perturbative correction(s). That's why we say that the perturbative and non-perturbative parts cannot be really separated in field theory (and string theory) except for cases when the whole (or whole except for a few terms) perturbative contribution vanishes. Unlike field theory, string theory tells us what the full result is. See also a text about non-perturbative well-definedness.

Now, when we have spent some time with clarifying what the adjective "perturbatively finite" means, let us ask:

What are the "finite amplitudes" of the supergravity theory that we often talk about? Well, they are the scattering amplitudes of the massless gravitons, graviphotons, scalars, gravitinos, and other fermions in the multiplet with 128 bosonic and 128 fermionic states.

At some level, you might say that these amplitudes encode all of physics. But it is not hard to see that physicists should look at other things, too. A theory with gravitons should include massive objects - because gravity's most favorite role is to couple to massive objects. In a similar fashion, supergravity contains 28 U(1) fields i.e. 28 copies of a "photon". Whenever you have photons, you should also have charged objects because photons love to interact with them.

In fact, love is not the only justification of this statement. ;-) A theory that has gravitational and electromagnetic fields - such as N=8 SUGRA - also admits classical solutions that look like massive and charged black holes. It seems almost guaranteed, both theoretically and phenomenologically, that some of these massive and charged solutions must correspond to actual objects in the theory, at least if the dimensionless charges are much greater than one and the classical limit is legitimate.

Let me assume that you believe me. If there are 28 copies of the electromagnetic field, there should exist all kinds of "electrons" that carry charges under those 28 U(1) groups. Because of the exceptional symmetry of the supergravity, electric and magnetic charges should be treated on equal footing. If electric charges exist, magnetic monopoles associated with the U(1) groups should exist, too.

To summarize: the space of allowed charges should be R⁵⁶.

Actually, this is just the classical result, describing charges of allowed black hole solutions and assuming that the charges are large enough for quantum mechanics to be irrelevant. Quantum mechanically, you can easily see that charges cannot be continuous. A magnetic monopole requires a Dirac string for the gauge field(s) to be well-defined and the wave functions of electrically charged objects must be single-valued when these particles rotate around the Dirac string.

This comment implies that the charges must be quantized and the unit of the magnetic charge is inversely proportional to the unit of the electric charge: that's known as the Dirac quantization rule. At any rate, the charges cannot live in the continuous 56-dimensional space. They must live in a 56-dimensional lattice instead. This conclusion follows by "pure thought" from gedanken experiments involving the Dirac strings and other objects. There is no way to avoid the conclusion.

Quantum mechanics often forces quantities to take discrete values: that's why it's called quantum mechanics. Get used to it. ;-)

Now, when you study how the 56-dimensional lattice (inside the 56-dimensional continuum we mentioned above) can look like, you find out that there are many possible lattices to do the job. Some elementary electric charges can be higher (and their associated magnetic charges must be lower) and the charges can mix with each other (the lattice can be tilted). If you study the space of possible solutions (56-dimensional lattices where everything works, including some Chern-Simons terms), you find out that the space is

E₇₍₇₎(Z) \ E₇₍₇₎ / SU(8),

the same 133-63 = 70-dimensional moduli space we know from string theory. Different lattices - different points on the moduli space - give non-equivalent theories because they give you different spectra of the allowed charges. Only if you map the lattice onto itself, you obtain the same background: that's why we have to take the quotient by the discrete group on the left side. The other groups already occur in classical supergravity.

Note that this non-equivalence of the vacua on the moduli space follows from quantum mechanics because quantum mechanics forced the charges to be quantized.

This has also reduced the original E_{7(7)} continuous symmetry of the classical supergravity theory to E_{7(7)}(Z), its important discrete subgroup (that we know as the U-duality group). You should see that any consistent quantum theory - where wave functions are single-valued - only allows the discrete, E_{7(7)}(Z) symmetry. The choice of the lattice is then physical. In the classical theory, charges are continuous, the symmetry group is also continuous, and all choices of the "continuous lattice", R^{56}, are equivalent. The absolute values of scalars are unphysical, only the relative ones matter.

With quantum mechanics, it can't be the whole story. Various things have to become quantized, new states have to appear, the scalars have to become physical parameters of the "environment". Supergravity without the new stringy effects is in the swampland, as argued by Green, Ooguri, Schwarz (2007) and others. See also Non-decoupling of N=8 d=8 SUGRA.

For the scattering of gravitons, graviphotons, and their supersymmetric friends, this quantization doesn't seem to matter perturbatively. The perturbative expansion effectively assumes that the charged particles cannot run in the loops at all. All the contributions with charged stuff in the loops is erased.

However, if you want the exact results, even for the scattering amplitudes of massless gravitons, it is important to realize that charged objects can run in the loops, too. Because these charged objects are massive, their pair production is exponentially suppressed (because in field theory, such a pair production of charged black holes would correspond to an instanton). Because of similar subtle technical reasons, the fact that the perturbative SUGRA completely misinterpreted the spectrum of allowed electromagnetic charges didn't spoil the perturbative result for the scattering amplitudes.

But non-perturbatively, you need to know what objects can run in the loops. Pair production of charged objects in the loops must be allowed. It adds exponentially suppressed terms and the quantization rules (the lattice) starts to matter. Non-perturbatively, the full stringy answer is the only consistent one and there is a 70-dimensional space of possible backgrounds (each of them giving you a superselection sector of the stringy/M-theoretical Hilbert space).

Whether the first inconsistencies in ordinary N=8 supergravity occur at the three-loop level, nine-loop level, or non-perturbative level is a technicality. What's qualitatively important for the "big picture" is that the N=8 supergravity cannot be the whole story. And it is, in fact, possible to deduce the aspects of stringy physics that have to be a part of the picture and that the old understanding of the N=8 supergravity neglected.

Extra dimensions

You find one more related surprise - one that has been known since the birth of supergravity theories but whose importance was not fully appreciated. The N=8 supergravity is the dimensional reduction of the 11-dimensional supergravity. If you take the "most beautiful" gravitational effective field theory in four dimensions, you are pretty much guaranteed to discover its 11-dimensional origin - another hint of string/M-theory. The 11-dimensional interpretation is clearly natural and more symmetric but the relationship is actually not about naturalness only: it is much more inevitable.

In classical field theory, dimensional reduction means that you assume that fields don't depend on certain dimensions at all and some corresponding components of vectorial and tensorial fields are set to zero. Consequently, the dimensions cease to exist. You kill them by hand. That's how Kaluza originally reduced the 5-dimensional Einstein's gravity to a four-dimensional electrogravitational theory.

In the full quantum theory, dimensional reduction is more subtle. As Klein realized, you must actually compactify the unwanted dimensions on circles. If the circles are very short, fields are not allowed to depend on the circular coordinate much because such variable modes would have a huge momentum (and energy). However, in the full theory, these variable (Kaluza-Klein) modes must still exist.

In the case of N=8 supergravity, it is the dimensional reduction of 11-dimensional supergravity or type IIA or type IIB supergravity in 10 dimensions. At the quantum level, when you require all quantization rules to hold, the theory arises from the compactification of M-theory on a seven-torus or type IIA or type IIB string theory on a six-torus.

The extra dimensions are not an additional or arbitrary assumption. I can always use Fourier series to reconstruct the additional dimensions from the wave functions of charged fields that already existed before we learned about the extra dimensions (charged black holes etc.). The only new thing is that such an approach turns out to be useful and it leads to new insights and a more concise theory.

Note that there are only six "elementary" supersymmetric string backgrounds in the highest (10 or 11) dimensions - M-theory; type IIA; type IIB; type I; heterotic E; heterotic O string theories. Because the N=8 supergravity is directly related to three of them (all the backgrounds with 32 supercharges, the maximum number), it is damn important in the structure of string/M-theory.

The maximally supersymmetric vacua of string theory - those related to the N=8 supergravity - are the best understood ones and a statement that you should avoid string theory exactly in this "inherently stringy" and special (and fully mapped!) segment of the landscape would be entirely ridiculous. The "mini-landscape" of the maximally supersymmetric backgrounds has been fully mapped, its (not only) qualitative physics has been entirely understood (since 1995, it's the string theorist's alphabet that he can recite in both directions), and we know for sure that all the stringy conclusions about any similar theory with 32 supercharges are inevitable.

Incidentally, the only other allowed set of vacua with 32 supercharges - besides M-theory on tori - involve type IIB string theory on AdS5 x S5 (or its orbifold), the best known example of the AdS/CFT correspondence. (In the N=4 gauge-theoretical "boundary CFT" description, 1/2 of the supercharges emerge as superconformal generators.) The more supersymmetry we have, the more we know about the internal structure of the theory.

I have explained that the consistent completion of supergravity has the same discrete charges and discrete symmetries as string/M-theory. But you can actually derive other stringy features of physics in the same way. For example, you can see that there are the strings in the game and they become damn important in a certain limit. You don't have to add the strings by hand: you can derive their existence and their importance by general considerations involving any theory similar to the N=8 SUGRA.

For example, the 70-dimensional space (U-duality\E_{7(7)}/SU(8)) of possible backgrounds (lattices) that we described above - a space that already existed in classical SUGRA as the configuration space of the scalar fields - has various limits.

In one limit, 7 charges become nearly continuous. This is the "decompactification" limit where 7 dimensions become very large and therefore useful: the charges are interpreted as momenta along these 7 directions. The local physics of the theory can be seen to be 11-dimensional supergravity. You may even find a lot of non-stringy evidence that these 11 dimensions must locally respect the Lorentz symmetry - they are parts of the same geometry; this proposition is clearly a fact once you adopt string theory. This theory has classical M2-brane and M5-brane solutions. Quantization arguments similar to those above imply that the charges of these M2-branes and M5-branes are discrete, too. Of course, the quanta are completely fixed and known in string/M-theory.

Take the minimum-charge M2-brane and wrap it on a circle: you produce a type II string. If this (wrapped) circle shrinks to a small size, and such a point is guaranteed to exist on the moduli space of the (any) "quantum-completed-supergravity", this string becomes the lightest degree of freedom. It can be shown to have the same interactions and importance as it is known to have in string theory: it has to be the same string, after all.

You can prove that it has certain internal degrees of freedom and they can be quantized to obtain the usual stringy tower of states (spectrum). Adding interacting strings that vibrate in the graviton "dance" can be shown equivalent to a modification of the background geometry, proving that the geometry is really made out of these strings in this limit, and so on.

If you think about physics carefully and if you have the required background in physics up to quantum field theory (that's enough!), you should be able to think in the "stringy way" in a few minutes. There are no "really new" arguments that wouldn't have been successfully tested in older physics disciplines (they are only put together in new ways to learn new fascinating things) - which is why Barton Zwiebach could have written an undergrad textbook of string theory and present it as a set of exercises about classical and quantum mechanics and electromagnetism. Afterwords, you will be using pretty much the same arguments as string theorists to deduce and calculate other detailed features of physics.

In some sense, you are still studying the N=8 supergravity. But you know much more about it. You need string theory to understand its non-perturbative physics properly. Again, perturbative string theory is only sufficient to crack a larger but still limited realm of questions. But we know a lot about non-perturbative string theory, too: most of the string research since the mid 1990s was dedicated to non-perturbative string theory.

Dualities, M-theory, detailed black hole thermodynamics, and partly also the AdS/CFT holography are among the most important concepts that have emerged in this research. It may be fair to say that this non-perturbative knowledge about string theory extends the previous knowledge about perturbative string theory in an analogous way as perturbative string theory itself extended the insights that can be obtained from ordinary supergravity (an effective quantum field theory). In this process, new things inevitably emerge such as new physical objects such as fields, particles, and branes, new quantization rules, new interactions, new symmetries and equivalences, new phase transitions, new limits, and new effective degrees of freedom that govern these limits.

We shouldn't really be using the term "string theory" for the full non-perturbative structure we have in mind today (strings are no longer "the fundamental objects") - except that the term "string theory" is convenient and it stuck. ;-)

We are still looking at the "same" theory as the N=8 SUGRA but many mysteries have evaporated and our understanding is much more detailed. It is simply not true that everything goes. 32 supercharges and physical consistency hugely constrain the laws of physics and string/M-theory on tori plus type IIB on AdS5 times a sphere (or orbifold) give the only solutions to these constraints.

Summary: consistency completes SUGRA to string/M-theory

To summarize, if you accept that the N=8 supergravity is on the right track, gedanken experiments and consistency inevitably lead you to the full string/M-theory as the only completion of the supergravity theory that is non-perturbatively consistent.

All people who studied supergravity in the early 1980s realize this fact today - they have understood that string theory is the only broader justification for their approximate calculations - and the supergravity community has become a subset of the string theory community (or stringy-supergravity federal community, if you want me to be excessively polite), especially once the importance of M-theory was appreciated, even though some people are still better at computing low-energy results and other people are better at other things.

Everyone who wants to "unlearn" these things is completely deluded. String theory has become an unavoidable component of a proper analysis of N=8 supergravity and many similar theories. It is known to be fully consistent while supergravity with continuous non-compact symmetries has been shown (above and elsewhere) to be non-perturbatively inconsistent.

Finally, I want to address a particular related myth that is being propagated in some not-really-serious but mildly influential corners - namely the statement that string theory is only known to be consistent perturbatively, much like supergravity. This is clearly false. When we talked about the lattices of charges and the (missing) loops with charged (and thus massive) stuff in supergravity, it was an explanation that supergravity was non-perturbatively inconsistent per se.

In string theory, these particular inconsistencies are absent, even non-perturbatively, and one can actually show that other inconsistencies are also absent. String theory properly describes even the most characteristic features of quantum gravity such as the black hole microstates and their evaporation: that's why string theorists could have computed the entropy of many black holes so much more accurately than general relativity. This fine, quantum analysis of black holes in string theory surely requires one to use non-perturbative tools.

But on the other hand, D-branes are an example of non-perturbative (very heavy, solitonic) objects that is nevertheless studied by (almost) the very same methods as perturbative string theory itself (with different boundary conditions etc.). A description of physics involving D-branes is "trans-perturbative" but it is effectively enough to get to any coupling, including infinite coupling. For example, black holes in strongly coupled stringy backgrounds can be mapped to fully understood configurations of D-branes and strings.

Also, perturbative string theory has a different expansion parameter than supergravity. String theory has a dimensionless coupling constant related to the dilaton (a scalar field) while supergravity has dimensionful Newton's constant that must be multiplied by a power of energy to become dimensionless. In the case of supergravity, the regime where the parameter goes to infinity is a high-energy regime (black hole production) that is not understood and looks inconsistent in (perturbative) SUGRA: the divergences of the Taylor expansion spoil the party and the perturbative definition gives you no rule how to resum the divergent series.

On the other hand, in string theory, when the coupling is sent to infinity, dualities allow us to use another, equivalent, and well-behaved description. In string theory, it's still impossible to resum the (still divergent) perturbative expansion but you can discover other objects that make the theory work beyond the perturbative series and they actually allow you to calculate the infinitely coupled limit exactly.

So in string theory, physics is clearly well-behaved for a small coupling, including the whole perturbative expansion and the non-perturbative consistency checks that failed in SUGRA. It also satisfies the very same consistency checks when the coupling goes to the other extreme, namely to infinity: SUGRA fails completely when you push it to the opposite limit. If you still think that this is insufficient to believe that string theory is also well-behaved everywhere in the middle, let me finally say that the maximally supersymmetric backgrounds - M-theory and its compactifications (from 0-torus to 5-torus) - can be fully and non-perturbatively defined in terms of Matrix theory (valid for any size of the torus and any energy and involving no expansions needed for the definition) which eliminates the remaining doubts.

Phenomenological road

In the first portion of the text, I have described the first, "consistency" road that inevitably leads everyone interested in supergravity to string theory. As we hinted at the beginning, the other road is based on phenomenology - the requirement that the theory agrees with the real world in detail and incorporates the Standard Model or its generalization.

The same N=8 supersymmetry that guaranteed those nice cancellations in the supergravity theory is also a bad thing. It is far too constraining. For every boson in your theory, it predicts a fermion - in fact, a lot of other fermions (and bosons!), too. All of their masses and interactions are forced to be "the same". On the other hand, the Standard Model (reality) describes fields and particles that are far more generic and unrelated. From this viewpoint, the extended supersymmetry looks like green shackles.

You might propose to break the N=8 supersymmetry spontaneously. However, N=8 supersymmetry cannot be spontaneously broken to N=1 or N=0 supersymmetries - the latter being two realistic choices. In fact, even N=2 supersymmetry is too constraining and cannot be spontaneously broken to N=1 or N=0, at least not by field theoretical methods in four dimensions.

If you can't break the N=8 supersymmetry spontaneously, you may want to break it explicitly. However, in that case, you lose the cancellations completely. Even with a reduced degree of supersymmetry in four dimensions (e.g. in other supergravity theories, including gauged supergravities), the divergences start to reappear, even in the perturbative expansion. No other gravitational field theories that are perturbatively finite, besides the N=8 supergravity, are known. (It is likely that even if such hypothetical additional theories existed, you would probably need string theory to find them and/or to prove the perturbative finiteness.)

So how can you preserve the nice features of the N=8 supersymmetry and get rid of the bad features? There is a solution. You may spontaneously (read this word carefully, Daniel from Sao Paolo!) break the maximally extended supersymmetry but you must do it in a non-field-theoretical fashion. In fact, you need to return to the full description of the N=8 supergravity which is M-theory on a seven-torus.

The correct spontaneous breaking of the supersymmetry instructs you to compactify M-theory or string theory on a different, more complicated manifold such as a G2 holonomy manifold (or a Calabi-Yau manifold times a line interval, or related choices). Without a loss of generality, a conceptual discussion may focus on the G2 holonomy manifolds.

Locally, G2 holonomy manifolds are made out of the same 7-dimensional space as 7-tori. In the full configuration space of string/M-theory, you may "connect" G2 manifolds and 7-tori. For this particular pair of manifolds, it can only be done off-shell - you need to consider configurations with non-zero energy as intermediate points in order to achieve this brutal topology change - but in some moral sense, G2 manifolds can be obtained by adding a vev to 7-tori.

The ultraviolet behavior is unaffected because at very short distances, 7-tori and G2 manifolds look the same: they are 7-dimensional manifolds, after all. In fact, "locally", there are still all 32 supercharges. So if M-theory on 7-tori is finite, it is not shocking that M-theory on G2 manifolds is finite, too. But let me warn you: for general backgrounds, the finiteness depends on string theory. If you truncate the stringy description into the massless fields, the perturbative finiteness of the field theory will never work as well as it did in the N=8 SUGRA, as far as I can say.

M-theory on G2 manifolds leads to N=1 supersymmetry in four dimensions which is a realistic degree of supersymmetry, one that can be spontaneously broken to N=0 supersymmetry. It describes a lot of phenomenologically acceptable models, including models that contain pretty much pure MSSM at low energies. So if you try to preserve as many "nice" aspects of the N=8 supergravity as you can get, it is actually possible to preserve the "full, maximally supersymmetric local physics" of the N=8 SUGRA, but in 11 dimensions rather than 4, and use a different space than a 7-torus for compactification. At any rate, you end up with the conventional superstring model building (and you may find other realistic classes of vacua that are a priori as OK as the G2 manifolds once you fully adopt the stringy rules of the game).

In the G2 context, at high energies (at distances shorter than the size of the G2 manifold), you still "morally" have 32 supercharges because the theory is M-theory, after all. At lower energies (at distances longer than the size of the G2 manifold), these 32 supercharges are broken down to 4 supercharges by the G2 geometry. Breaking of symmetry at long distances (or low energies) is exactly what you are used to from field theory except that string theory and its extra dimensions gave you completely new tools to achieve this goal.

The set of a priori possible vevs giving us semi-realistic compactifications is rather large - it is the landscape - but if you realize how unconstrained and "arbitrary" the real world (and its particle spectrum) is in comparison with the N=8 supergravity (where the mini-landscape is very simple), you should understand that the large size of the N=1 or N=0 landscape is another "required" feature that qualitatively agrees with observations. Whether there exists a "nearly unique" way to find the correct set of vevs - or whether the anthropic people are correct in saying that we live in a "random", trash Universe - remains to be seen. But again, the knowledge that the landscape exists is another step that deepens our understanding of physics.

Once you realize most (not necessarily all) of these basic physical arguments, there is no way for you to "unrealize" them. String/M-theory describes the only non-perturbatively consistent completion as well as the only phenomenologically acceptable generalization of "nice" theories such as the N=8 supergravity and I would like to claim that a physically literate, intelligent reader can find the proof in this very essay. If you think that the N=8 supergravity is on the right track and you try to make the picture more complete, more self-consistent, or more compatible with observations, you are led in the same direction in all three cases: you are led to string/M-theory.

Everyone who pretends that it is not the case is deluded. There is no way for theoretical physics to "forget" string/M-theory because it has answered many questions about older theories such as the N=8 supergravity that used to be puzzling but that cannot be "unanswered" today.

Entertainingly enough, one of Peter Woit's passionate readers, enthusiastic creationist Roger Schlafly, was more self-consistent than Peter Woit when he criticized Woit's attempts to promote the N=8 supergravity as a separate entity (and a full-fledged candidate for a unifying theory):

Peter, you are going soft on us. This is just another wacky theory with no connection to reality. It does not because [LM: should be: "become"] valid or useful just because some of the infinities cancel.

Of course, when your brain works properly, there is no way to "cut the strings" away from the rest of physics. Such an attempt is fully analogous to the creationist attempts to divide the evolution into the "naturally tolerable one" and "one that surely requires God". There exists no point where you could draw such a boundary (read about the unity of strings). In fact, when you ask different creationists where this hypothetical boundary is supposed to be located, they will pinpoint all possible points. ;-)

The very same observation applies to the anti-string-theoretical crackpots: read more about the evolution-string-theory analogies. When you ask them when physics exactly began to generate wrong results, they will give you all possible years and all possible boundaries between papers.

Peter Woit puts the boundary somewhere in the early 1980s because this is when he realized that his brain was not capable to do physics. Lee Smolin places the boundary somewhere in the 1970s (or earlier) because this is when this particular critic of science realized the same thing - that's also why Smolin tells you that all of particle physics sucks, in a sense - while Roger Schlafly puts the boundary somewhere to the 19th century because this is when physics began to support things like the old Earth and Darwin's theory that he simply cannot stand. ;-)

The only semi-consistent solution for anti-scientific activists such as Peter Woit would be to deny all of physics completely. Science hasn't stopped (and it will never end) and even after 1838 or 1975 or 1982, there exists a significant portion (thousands) of completely solid physics papers. Any attempt to draw a boundary somewhere in the middle of physics can be proven ludicrous because physics is a very tightly connected conglomerate of insights and evidence. An attempt to cut physics into two pieces along an argument can be shown illegitimate by repeating the proof that the argument is actually correct.

For example, an attempt to say that the N=8 supergravity with the non-compact continuous E_{7(7)} symmetry was still on the right track while its stringy completion with the discrete U-duality symmetry is not contradicts the single-valuedness of the wave functions, as used in the Dirac quantization argument. It contradicts hundreds of other facts about physics, too, but I wanted to be very specific.

String theory is indisputably at the very heart of physics which is at the very heart of natural science which also means that the critics of string theory find themselves at the very center of a gigantic excrement.

And that's the memo.