Un-integrated PDFs in CCFM

Quick read on Jung's DIS2004 proceedings. Bottom line: the modest claim is fine, the ambitious one isn't yet earned, and the paper is short enough that you can see both in ten minutes. Worth a look if you care about small-x gluon densities; not something to build on directly — cite Jung-Salam [1] for the machinery instead. What's actually new here is narrow but real: the alternative factorization scale mu_f = pt/(1-z) with the resulting sets A1 and B1, the mu_r variants A0-/A0+/B0-/B0+, and the side-by-side of evolved densities against their no-branching counterparts in Fig. 1. That last comparison is the useful part. The nobran curves are a clean decomposition of Eq. 1, and they do support the statement that intrinsic-kt influence is concentrated below roughly 1 GeV while above that the perturbative evolution dominates and quite different starting distributions converge. That convergence is shown, not asserted, and it's the paper's best moment. Where I'd push back on the reader's report: the circularity score of 5 is too harsh. Fitting a uPDF to F2 and then reporting the fit quality is what a PDF fit is; the non-circular content is the kt-dependence, which F2 does not constrain. The genuinely soft spot is the one the stress-test names, and it lands. Section 3 says outright that the choice of kcutt changes the small-kt distributions and that the visible dip in Fig. 1 is 'directly related to kcutt.' Then Table 1 varies k0 and kcutt together — A0 at (1.33, 1.33), B0 at (0.8, 0.25) — so nothing separates intrinsic width from regulator. The author is candid about the kcutt dependence, which I credit, but candour isn't a closure test. Until one knob is held fixed while the other moves, 'the small-kt region provides new information on the non-perturbative part' is a promissory note. Smaller things: Q0 is never stated, the collinear cutoff appears as both 1.3 and 1.33 GeV, and the mu_r variation is quoted after refitting p0 at each scale, so those chi2 values aren't an unabsorbed theory error. All fixable, none fatal. Who should read it: small-x phenomenologists and anyone using CCFM-based generators for forward-jet or forward-Drell-Yan estimates, mainly as a map of which knobs matter. It deserves referee time as a short note, with the kcutt/k0 degeneracy as the one comment that must be answered.

The central interpretive claim of §3 is that the small-kt region of the uPDF carries non-perturbative information, because above roughly 1 GeV the perturbative evolution dominates and different starting distributions converge. The evidence for the 'intrinsic' part is the nobran curves in Fig. 1 (the first term of Eq. 1), which is a fair decomposition and does support the statement that intrinsic influence is concentrated at small kt. The soft spot is the attribution of the low-kt shape to intrinsic physics rather than to the regulator. The paper states in §3 that 'the specific choice of the kcutt results in different distributions at small values of kt' and that 'the dip in kt visible in Fig. 1 (upper plot) is directly related to kcutt.' A visible structure in the very region claimed to carry non-perturbative information is thus acknowledged to be a cutoff artifact. More importantly, in Table 1 the two knobs that control the low-kt shape, kcutt and k0, co-vary across the sets: A0 has (k0, kcutt) = (1.33, 1.33), B0 has (0.8, 0.25), and no set holds one fixed while varying the other. The scan therefore cannot separate 'intrinsic width' from 'regulator'. The conclusion that low kt is a clean window on non-perturbative QCD is promissory: no observable is shown that actually constrains it, and the one visible feature is a cutoff effect. This does not invalidate the convergence claim at large kt, which is supported, but it does undercut the 'new information' conclusion.

The reader's weakest_assumption identifies the right load-bearing concern: the low-kt shape is partly a regulator choice, and the paper offers no closure test separating kcutt from intrinsic physics. I agree and would sharpen it in one respect. The paper does acknowledge the kcutt dependence, so per the review rules that acknowledged limitation is not a new finding; what remains unresolved is that the presented sets confound the two knobs. Sets A and B differ in both kcutt (1.33 vs 0.25 GeV) and k0 (1.33 vs 0.8 GeV), and no set varies one at fixed other, so the scan cannot attribute the low-kt structure to intrinsic width rather than cutoff. The acknowledged dip in Fig. 1 being 'directly related to kcutt' is the sharpest form of this: a visible feature sits in exactly the region claimed to be a non-perturbative window. The rest of the paper holds up for what it is. The convergence of different starting distributions at kt above roughly 1 GeV is supported by the nobran comparison in Fig. 1, and Table 1 documents which variants survive the F2 comparison. The renormalization-scale study is transparently a refit of p0 rather than an unabsorbed uncertainty band, which is a methodological weakness but not a flaw in the central claim. The Q0 omission and the 1.3 vs 1.33 GeV cutoff inconsistency are reporting issues that should be fixed but do not move the verdict. Because the reader is already at CONDITIONAL for essentially this reason, my read does not change the verdict; it supplies the specific one-knob scan that would settle whether the low-kt window claim survives. Confidence in the paper's modest claims is moderate; the low-kt interpretation should be softened until that scan is shown.

The paper's useful and checkable content is the decomposition of the uPDF into the non-resolved term of Eq. (2) and the evolved term of Eq. (1), shown as the 'nobran' curves in Fig. 1, and the observation that three quite different starting distributions (Gaussian sets A and B, and the GBW form) converge at kt above about 1 GeV while differing substantially below it. That is a falsifiable statement: it predicts that observables with sensitivity to the gluon kt distribution, such as dijet azimuthal correlations or forward-jet spectra at HERA, should discriminate among the sets where inclusive F2 cannot. The inclusion of a saturation-motivated starting distribution as an explicit alternative is a sensible cross-check, and the tabulation of fit quality for each variant is the right way to present such a comparison in a proceedings. The value of the paper is as a compact status report on the systematic freedoms in CCFM uPDF fits rather than as a new result.

The paper asks what the shape of the un-integrated gluon density at low transverse momentum $k_t$ actually tells us, and whether that information survives the freedom in how the evolution is set up. Working with the CCFM evolution equation, the author evolves several starting distributions — Gaussian forms with different widths and soft cutoffs, plus a saturation-motivated form — and fits each to the HERA structure-function data on $F_2(x,Q^2)$ at small $x$. The central observation is that all of these variants describe $F_2$ about equally well, yet they differ visibly at low $k_t$; above roughly $1$ GeV the perturbative evolution washes out the differences and the curves converge. The author's conclusion is that the low-$k_t$ region is a genuine window on the non-perturbative part of the parton density, and that inclusive measurements are blind to it. The paper also maps how the renormalization and factorization scales shift the fit parameters, showing that changing them changes the length of the evolution ladder and hence the required input.

What carries the argument. The carrier of the argument is the CCFM evolution equation in its integral form, $xA(x,k_t,bar{q}) = xA_0(x,k_t,bar{q}) + int (dz/z) int (dq^2/q^2) Theta(bar{q}-zq) Delta_s tilde{P}(z,q,k_t) xA(x/z,k_t',q)$, together with its first term $xA_0(x,k_t,Q_0)Delta_s(bar{q},Q_0)$, which encodes the probability of no resolvable radiation between the starting scale and the factorization scale. The splitting function used is the simple gluon one, $P_{gg}(z,bar{q},k_t) = bar{alpha}_s(q^2(1-z)^2)(1-z) + bar{alpha}_s(k_t^2) z Delta_{ns}(z,q^2,k_t)$, with the non-Sudakov form factor $Delta_{ns}$ resumming the $1/z$ behaviour in the region where $k_t$ is ordered against the emitted gluon. Angular ordering makes the propagator gluon $k_t$ perform a random walk that can reach very small values, which is why a soft cutoff $k_{cut t}$ must be introduced; that cutoff, together with the Gaussian width $k_0$ and the exponent $p_0$ of $xA_0(x,k_t,Q_0) = N x^{p_0}(1-x)^{p_1} exp(-k_t^2/k_0^2)$, forms the parameter set fitted to data.

Core claim. The paper's central claim is a separation of regimes inside a single object. In the un-integrated gluon density $xA(x,k_t^2)$ obtained from CCFM evolution, the influence of the intrinsic transverse-momentum distribution is confined to small $k_t$, while at $k_t$ above roughly $1$ GeV the perturbative evolution is what determines the shape. Because the evolution is angular ordered, the starting distribution is largely forgotten at larger $k_t$: the Gaussian sets A and B and the saturation-motivated GBW set, which differ substantially as inputs, produce similar evolved densities at large $k_t$. All of them nonetheless give a comparable description of the measured $F_2$ data, with $k_t$ cutoffs and Gaussian widths traded against each other in the fit. The author draws the consequence directly: the small-$k_t$ region is where new information about the non-perturbative part of the parton density lives, and that information is not recoverable from inclusive structure functions. The paper further reports that varying the renormalization scale in the off-shell matrix element requires refitting the input parameter $p_0$, and that moving the factorization scale definition from $muf = sqrt(hat{s}+Q_perp^2)$ to $muf = p_t/(1-z)$ also demands a different $p_0$ while still describing the data.

The premise it rests on. The claim that the low-$k_t$ shape carries real non-perturbative information assumes that this shape is not mostly an artifact of the soft cutoff $k_{cut t}$, which the author fits to data and which he himself notes produces the differing small-$k_t$ distributions and the visible dip in Fig. 1.

Measurements that resolve the low-$k_t$ gluon directly, rather than through $F_2$, become the only way to discriminate between Gaussian and saturation-motivated starting distributions.

Because sets with quite different inputs converge at $k_t$ above about $1$ GeV, describing $F_2$ well is not evidence that the correct non-perturbative input has been found.

The renormalization scale in the off-shell matrix element must be treated as part of the fit, not as an external choice, since changing it moves the required $p_0$.

Redefining the factorization scale from the quark-pair invariant mass to a single-quark quantity $p_t/(1-z)$ still permits a good description of the data, but only with a significantly different input parameter.

If the low-$k_t$ region is genuinely non-perturbative, saturation effects at small $k_t$ are directly addressable with un-integrated densities rather than inferred indirectly.

The paper's logic suggests a concrete discriminating programme: observables sensitive to the gluon $k_t$ spectrum, such as forward-jet or dijet azimuthal decorrelation at HERA and the LHC, should separate the Gaussian sets from the saturation-motivated GBW set even where $F_2$ cannot.

Since the visible dip in the low-$k_t$ distribution in Fig. 1 tracks the soft cutoff $k_{cut t}$, a clean closure test would hold the input shape fixed and vary only $k_{cut t}$, to see how much of the claimed non-perturbative structure is regulator rather than physics.

The convergence of very different inputs at large $k_t$ is a statement about the memory loss of an angular-ordered ladder; it suggests that the informative content of any un-integrated density is concentrated in the region where the ladder has not yet run, which is a general feature worth testing in other evolution schemes.

If low-$k_t$ gluon densities are the only handle on the non-perturbative input, then the current spread among fitted parameter sets is itself a quantifiable uncertainty band for small-$x$ predictions, and should be propagated into cross-section estimates rather than quoted as alternative tunes.

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review