WorksPhysics and Astronomy
Testing the Gaussian Ansatz for TMD PDFs Through CMS Measurements
Abstract
We investigate the differential cross section of the Drell-Yan process within the framework of Transverse Momentum Dependent (TMD) factorization. We perform phenomenological calculations of the differential cross section as a function of in different dilepton invariant mass bins and compare it with the measured data from the CMS experiment at TeV, employing two distinct transverse momentum dependencies of the parton distribution functions: the Collins-Soper-Sterman (CSS) evolution method and the Gaussian Ansatz. The results indicate that the CSS evolution method agrees well with experimental data across the entire energy range, demonstrating its strong predictive capability for the evolution behavior of TMD distribution functions. In contrast, the Gaussian Ansatz of the TMD PDFs is only applicable in the low-energy region and exhibits significant limitations at high energies. This study verifies the effectiveness of the TMD factorization theorem in describing high-energy processes and provides theoretical support for an in-depth exploration of the nucleon's internal structure and non-perturbative QCD dynamics.
Introduction
Transverse Momentum Dependent Parton Distribution Functions (TMD PDFs) serve as the core probes for exploring the three-dimensional internal structure of nucleons. Under the color confinement mechanism of Quantum Chromodynamics (QCD) [1, 2, 3], partons inside the nucleon not only carry longitudinal momentum but also exhibit transverse motion () and spatial distribution (). These dynamical behaviors directly encode the non-perturbative information of QCD in the low-energy region. By introducing the transverse momentum of partons, TMD PDFs map the three-dimensional partonic structure of the nucleon onto observables, thereby serving as a crucial bridge connecting fundamental QCD theory and experimental measurements of hadron structure. Therefore, a precise understanding of the transverse momentum dependence of TMD PDFs is of indispensable significance for revealing the complete image of quarks and gluons inside the nucleon.
The theoretical foundation for describing processes sensitive to transverse momentum was established through the development of the TMD factorization theorem. The original formalism by Collins, Soper, and Sterman (CSS) [4, 5, 6] provided a rigorous framework to factorize cross sections into hard functions, TMD PDFs, and soft factors, while systematically resumming large logarithmic terms via the Collins-Soper evolution equation. This formalism was later refined and extended, with key contributions from Collins in Ref. [7] and from Ji, Ma, and Yuan (JMY) [8]. These modern formulations, often expressed in Fourier space with respect to the impact parameter , introduced additional scale dependencies—the renormalization scale and the Collins-Soper scale —to properly handle rapidity divergences. The resulting factorization theorems are equivalent to those derived in Soft-Collinear Effective Theory (SCET) [9, 10, 11, 12] and form the cornerstone of modern TMD physics, enabling precise predictions for a wide range of processes including Semi-Inclusive Deep Inelastic Scattering (SIDIS) [8, 13, 14, 15, 16, 17], Drell-Yan (DY) dilepton production [7, 18, 19], and annihilation [20, 21] processes.
In early phenomenological studies, the transverse momentum dependence of TMDs was often modeled using a simple Gaussian Ansatz, assuming a factorized form as [22]. This Generalized Parton Model (GPM) [23, 24, 25] successfully described a wide range of low-energy data from experiments such as HERMES and COMPASS [24, 25, 26]. In this approach, the transverse momentum dependence of the cross section simplifies to a Gaussian form, where the observed transverse momentum of the final-state hadron in SIDIS process is governed by the sum of the widths from the TMD PDF and the fragmentation function. While this simple model has the advantage of being analytically tractable and providing a satisfactory description of data within specific kinematic cuts, it violates the full QCD evolution of TMDs. Modern TMD formalism predicts a more intricate behavior in which the transverse momentum distribution is controlled by the Sudakov-like form factor that resums large logarithms governs the scale dependence of TMDs. Consequently, the intrinsic transverse momentum distribution is no longer a fixed Gaussian but evolves with the hard scale, reflecting the underlying QCD dynamics.
Recently, the CMS Collaboration reported measurements of the double differential cross section for Drell-Yan process as functions of and the invariant mass of the final-state dilepton at TeV with an integrated luminosity of 36.3 fb [27]. It provides an ideal opportunity to assess how well these two distinct approaches—the simplified Gaussian parametrization and the full Sudakov evolution—describe the data across different kinematic regimes. In this work, we study the transverse-momentum spectrum of Drell-Yan dilepton within the TMD factorization framework. Two approaches to TMD evolution in the initial-state proton are considered, and the resulting predictions are compared with CMS data. This comparison allows us to probe the partonic structure of the proton, shedding light on its three-dimensional structure and nonperturbative QCD dynamics.
The remainder of this paper is organized as follows. In § 2, we present the theoretical calculation of the differential cross section for the Drell–Yan process in proton–proton collisions within the framework of TMD factorization. To incorporate TMD evolution effects, two approaches are employed: the CSS evolution formalism and the Gaussian Ansatz. In Sec. III, we provide numerical results, including theoretical predictions from both approaches and compare them with experimental data from the CMS experiment. Finally, Sec. IV summarizes the main findings of this work and presents our conclusions.
Framework
The process under study is the Drell-Yan process
where and denote the four-momenta of the two incident protons, and denotes the momentum of the virtual photon. In contrast to that in the SIDIS process, is a time-like vector in the Drell-Yan process, namely, , which is the invariant mass squared of the final-state lepton pair. In order to express the differential cross section of the process, we introduce the following kinematical variables
where is the center-of-mass energy squared; and are the light-front momentum fractions carried by the annihilating quarks in the two initial-state protons, respectively; is the longitudinal momentum of the virtual photon in the c.m. frame of the incident protons; is the Feynman variable, which corresponds to the longitudinal momentum fraction carried by the lepton pair; and is the rapidity of the lepton pair. Thus, the differential cross section for the Drell-Yan process described in Eq. (1) has the form [6]
where is the cross section at tree level. represents the transverse momentum of the virtual photon, while and denote the transverse momenta of the interacting quarks in proton 1 and proton 2, respectively. Performing the Fourier Transformation on the -function in Eq. (3), one can obtain the following form
Here, we neglect the energy dependence of the TMD PDFs, which will be discussed in detail in the next subsection.
The differential cross section of the Drell-Yan process using the CSS method
Since it is much more convenient to solve the TMD evolution equations in coordinate space (-space) than in transverse momentum space (), where is conjugate to via Fourier Transformation, the differential cross section is usually expressed as a -dependent function formulated according to TMD factorization. Then, it can be transformed back to transverse momentum space through inverse Fourier Transformation. The transformation of TMD PDFs from transverse momentum space to -space is given by
Therefore, the differential cross section in Eq. (4) can be rewritten as
The structure function contains all-order resummation results and dominates at low values, while the term provides the necessary correction at moderate values. In this work, we will assume , neglecting the -term, and only consider the first term on the r.h.s. of Eq. (6).
In general, TMD factorization aims to separate well defined TMD functions so that they can be used in different processes while including the scheme/process dependence in the corresponding hard factors. can be expressed as the convolution of the TMD distribution functions at -space and hard factor [28]
where is the subtracted distribution function in -space with the superscript “sub” representing that the soft factor is subtracted. is the factor associated with hard scattering, is the renormalization scale in the case of collinear PDFs, and denotes an energy scale related to the cutoff of the TMD distributions. We should note that the distribution function is universal. However, the way to subtract the soft factor in the distribution function depends on the scheme used to regulate the light-cone singularity in the TMD definition [5], whereas the hard factor is also scheme-dependent. In the literature, two different schemes are usually used: the Collins-11 scheme [7] and the Ji-Ma-Yuan scheme [29]. If we perform a Fourier Transformation on , we obtain the distribution function in the transverse momentum space, which contains information about the probability of finding a quark with the collinear momentum fraction and transverse momentum .
The TMD evolution for the dependence of TMD PDFs is encoded in a Collins-Soper (CS) [7] equation through
while the dependence is derived from the renormalization group equation as
where is the CS evolution kernel, and and are the anomalous dimensions. The solutions of these evolution equations are studied in detail in Ref. [7]. Here, we will only discuss the final result. The overall solution structure for is the same as that for the Sudakov form factor. Namely, the energy evolution of TMDs from an initial energy to another energy is encoded in the Sudakov-like form factor through the exponential form . Hereafter, we set and express as for simplicity
where is the hard factor and depends on the scheme that we choose.
The -dependence of TMD distributions determines the transverse momentum dependence of experimental observables through Fourier Transformation, which makes understanding the -dependence important. In the small region, the dependence is perturbatively calculable, while at large the dependence becomes nonperturbative and should be obtained from experimental data. To combine the perturbative information at small with the nonperturbative part at large , a matching procedure must be introduced with a parameter serving as the boundary between the perturbative and nonperturbative regions. A -dependent function is defined, which has the property at low values of and at large values. A typical value of is chosen around , such that is always in the perturbative region. There are several different prescriptions in the literature [30, 31]. In this work, we adopt the original prescription introduced in Ref. [7], with .
In the small region , the defined TMDs at fixed energy can be expressed as a convolution of perturbatively calculable hard coefficients and the corresponding collinear PDFs [5, 32]
where stands for the convolution in the momentum fraction
and is the corresponding collinear PDF of the flavor in hadron at the energy scale , which can be a dynamic scale related to by , with and being the Euler Constant [5]. In addition, the sum runs over all parton flavors. Independent of the type of initial hadrons, the perturbative hard coefficients have been calculated for the parton-target case [33, 34] and the results have been presented in Ref. [32] (see also appendix A of Ref. [33]) as
where , , and is the strong coupling constant at the energy scale . The prescription prevents from hitting the so-called Landau pole in the large regime. We note that the -coefficients do not depend on the TMD scheme and on the initial hadrons. Namely, the -coefficients in Eq. (14) are universal for different schemes and initial hadrons.
The Sudakov-like form factor in Eq. (11) can be separated into a perturbatively calculable part and a nonperturbative part
The perturbative part of has the form
The coefficients and in Eq. (17) can be expanded as series:
In this work, we take to and to up to NLL order (with and ) [6, 16, 33, 35, 36, 37]:
For the nonperturbative form factor in the Drell-Yan process, a general parameterization associated with the TMD PDF has been proposed in Ref. [38] and it has the following form:
In Ref. [38], the parameters , , and are fitted from the nucleon-nucleon Drell-Yan process data at the initial scale with , , and . With the extracted parameters, the fitted results are , , and .
Since the nonperturbative form factor for quarks and antiquarks satisfies the following relation [39]
and the quark and antiquark contributions to are the same:
the associated with the TMD distribution function of one of the initial protons can be expressed as
With all the ingredients presented above, we can rewrite the -space distribution function
One should notice that the hard factors , which appear in Eq. (27) and in Eq. (7), are scheme dependent. In the two different schemes, the Collins-11 scheme [7] and the Ji-Ma-Yuan scheme [8, 29], the hard factors and have different forms:
where is another parameter to regulate the light-cone singularity of TMD distributions. However, if one absorbs both of the hard factors and into the definition of the -coefficients, the -coefficients turn out to be -independent (scheme-independent) and can be expressed as [40]
Substituting the above new -coefficients from Eq. (32) and Eq. (33) into Eq. (12), we can obtain
Finally, the structure function in -space can be written as
Thus, substituting Eq. (35) back into Eq. (4), one can obtain the final expression for the differential cross section as
The differential cross section for the Drell-Yan process using the Gaussian Ansatz
From Eq. (3), it can be inferred that the differential cross section for the Drell-Yan process has the following form
The transverse momentum dependence of the unpolarized distribution functions and can be assumed to follow a Gaussian-like distribution as
where represents the mean value of the squared transverse momentum of the partons in the initial-state protons. Substituting Eq. (38) and Eq. (39) into Eq. (37), we can obtain
After further calculations, the following expression of the differential cross section is obtained
NUMERICAL ESTIMATE
In this section, we will perform the numerical calculation utilizing both methods and compare the results with the experimental data from the CMS Collaboration.
Regarding the CSS method, we adopt the NLO set of the CT10 [41] parametrization (central PDF sets) for the collinear unpolarized distribution function of the proton as input. The values of the strong coupling are consistently obtained in 2-loop order as
with fixed and GeV. The running coupling in Eq. (42) satisfies .
The CMS Collaboration has measured double-differential cross sections of Drell–Yan dilepton production as a function of the invariant mass and transverse momentum at TeV, using data collected in 2016, corresponding to an integrated luminosity of 36.3 fb [27]. It provides us with an ideal opportunity to compare the validity of two different methods for describing the TMD dependence of the PDFs.
The differential cross sections obtained from the two different calculation methods are both four-dimensional differential cross sections with respect to , , and . However, the CMS Collaboration measured the cross section as a function of for different invariant mass bins: . In order to be consistent with the experimental data, we integrate over the remaining two variables within the kinematical range of CMS and the corresponding invariant mass bins. The integration variables and their ranges corresponding to the kinematic coverage of the CMS data are as follows
As a comparison with experimental data, the following ranges of have been adopted for the integration over : 50 GeV 76 GeV, 76 GeV 106 GeV, 106 GeV 170 GeV, 170 GeV 350 GeV, and 350 GeV 1000 GeV. The integration range for is from -2.4 to 2.4.
For dilepton invariant masses above the -boson mass, the contribution from -boson production becomes significant, and the above expressions must be modified accordingly. Therefore, except for the first interval GeV 76 GeV, we replace the quark electric charge with the electroweak charge for the other intervals, as follows:
The electroweak couplings are given in terms of the Weinberg angle by
with , GeV, and GeV from the PDG [42]. The couplings and are for up and down type quarks, respectively.
The numerical results are shown in Figure 1, where the red data points represent the experimental measurements with errors, and the solid black line corresponds to the theoretical calculations obtained using the CSS method. From Figure 1, we can see that the theoretical predictions based on the CSS resummation describe the general shape of the data well over wide ranges of and . The comparison demonstrates that the CSS resummation framework captures the essential features of the spectrum across a broad range of invariant masses.

Figure 1. The distribution of the differential cross section using the CSS method as a function of the lepton pair transverse momentum is presented for the lepton pair invariant mass ranges of 50 GeV76 GeV, 76 GeV106 GeV, 106 GeV170 GeV, 170 GeV 350 GeV and 350 GeV 1000 GeV.
The numerical estimate of the differential cross section from the Gaussian Ansatz with the experimental measurement is shown in Figure 2. From the figure, one can conclude that the Gaussian Ansatz of the TMD PDFs cannot describe the experimental data at high energy scales. The results of the structure functions utilizing the CSS method and the Gaussian Ansatz are shown in Figure 3. The figure shows the structure function of the CSS and the structure function of the Gaussian Ansatz as a function of the lepton pair transverse momentum at and , for energies GeV, GeV, GeV, and GeV. The structure functions are as follows:

Figure 2. The distribution of the differential cross section using the Gaussian Ansatz as a function of the lepton pair transverse momentum is presented for the lepton pair invariant mass ranges of 50 GeV–76 GeV, 76 GeV–106 GeV, 106 GeV–170 GeV, 170 GeV–350 GeV and 350 GeV–1000 GeV.

Figure 3. Setting and , the distribution of the structure function as a function of was calculated for GeV, GeV, GeV and GeV. The red solid line represents the distribution under the CSS approach, while the black solid line corresponds to the result from the Gaussian method.
In the figure, the dashed line represents the structure function calculated using the CSS method, while the solid line corresponds to the results from the Gaussian method. It can be observed that at low energies, the results from the Gaussian method agree well with those from the CSS method. However, as the energy increases, the Gaussian method begins to deviate from the CSS results and decreases rapidly. Therefore, it can be shown that the Gaussian method is only compatible with the CSS results in low-energy scenarios, which can describe the experimental data well over a wide energy range. Therefore, we conclude that the CSS formalism provides a reliable description of the TMD dependence of TMD functions across a broad range of energy scales, while the Gaussian Ansatz breaks down at high energy scales.
conclusion
In this work, we have studied the transverse momentum dependent differential cross section in the pp Drell-Yan process and compared our predictions with the CMS data collected in 2016, with an integrated luminosity of 36.3 fb. Within the TMD factorization framework, the calculation utilized two distinct ways to deal with the transverse momentum dependence of TMD PDFs. The first strategy is the Collins-Soper-Sterman (CSS) resummation formalism, where the evolution from an initial scale to a hard scale is encoded in a Sudakov form factor, . The second strategy is a Gaussian-type ansatz for the transverse-momentum dependence.
The numerical comparisons show that the CSS resummation formalism provides a consistent description of the measured spectra over a wide range of invariant masses, while the Gaussian Ansatz is only compatible with CSS expectations at relatively low and breaks down at higher scales. These results support the applicability and predictive power of TMD factorization with CSS evolution for collider Drell-Yan observables, which help us better understand the scope of applicability and predictability of the TMD factorization theorem. In the future, theoretical calculations based on TMD factorization with CSS evolution can be employed to predict or validate the measurements from collider experiments to further constrain the three-dimensional partonic structure of the nucleon.
Acknowledgments
This work is supported by the Natural Science Foundation of Henan Province under Grant Nos. 242300421377 and 232300421140.
References
References
- [1]D. J. Gross and F. Wilczek, Phys. Rev. Lett. 30, 1343 (1973).
- [2]H. D. Politzer, Phys. Rev. Lett. 30, 1346 (1973).
- [3]D. J. Gross and F. Wilczek, Phys. Rev. D 8, 3633 (1973).
- [4]J. C. Collins and D. E. Soper, Nucl. Phys. B 197, 446 (1982).
- [5]J. C. Collins and D. E. Soper, Nucl. Phys. B 193, 381 (1981), [Erratum: Nucl.Phys.B 213, 545 (1983)].
- [6]J. C. Collins, D. E. Soper, and G. F. Sterman, Nucl. Phys. B 250, 199 (1985).
- [7]J. Collins, Foundations of Perturbative QCD, vol. 32 (Cambridge University Press, 2011), ISBN 978-1-009-40184-5, 978-1-009-40183-8, 978-1-009-40182-1.
- [8]X.-D. Ji, J.-P. Ma, and F. Yuan, Phys. Rev. D 71, 034005 (2005).
- [9]C. W. Bauer, S. Fleming, and M. E. Luke, Phys. Rev. D 63, 014006 (2000).
- [10]C. W. Bauer, S. Fleming, D. Pirjol, and I. W. Stewart, Phys. Rev. D 63, 114020 (2001).
- [11]C. W. Bauer and I. W. Stewart, Phys. Lett. B 516, 134 (2001).
- [12]C. W. Bauer, D. Pirjol, and I. W. Stewart, Phys. Rev. D 65, 054022 (2002).
- [13]S. M. Aybat and T. C. Rogers, Phys. Rev. D 83, 114042 (2011).
- [14]J. C. Collins and T. C. Rogers, Phys. Rev. D 87, 034018 (2013).
- [15]X.-D. Ji and F. Yuan, Phys. Lett. B 543, 66 (2002).
- [16]M. G. Echevarria, A. Idilbi, A. Schäfer, and I. Scimemi, Eur. Phys. J. C 73 (2013).
- [17]M. G. Echevarria, A. Idilbi, A. Schäfer, and I. Scimemi, Eur. Phys. J. C 73, 2636 (2013).
- [18]D. Boer and W. Vogelsang, Phys. Rev. D 74, 014004 (2006).
- [19]S. Arnold, A. Metz, and M. Schlegel, Phys. Rev. D 79, 034005 (2009).
- [20]D. Pitonyak, M. Schlegel, and A. Metz, Phys. Rev. D 89, 054032 (2014).
- [21]D. Boer, Nucl. Phys. B 806, 23 (2009).
- [22]C. W. Gardiner and D. P. Majumdar, Phys. Rev. D 2, 2040 (1970).
- [23]M. Anselmino, M. Boglione, U. D’Alessio, A. Kotzinian, F. Murgia, and A. Prokudin, Phys. Rev. D 71, 074006 (2005).
- [24]P. Schweitzer, T. Teckentrup, and A. Metz, Phys. Rev. D 81, 094019 (2010).
- [25]M. Anselmino, M. Boglione, J. O. Gonzalez Hernandez, S. Melis, and A. Prokudin, JHEP 04, 005 (2014).
- [26]J. Cammarota, L. Gamberg, Z.-B. Kang, J. A. Miller, D. Pitonyak, A. Prokudin, T. C. Rogers, and N. Sato (Jefferson Lab Angular Momentum), Phys. Rev. D 102, 054002 (2020).
- [27]A. Tumasyan et al. (CMS), Eur. Phys. J. C 83, 628 (2023).
- [28]A. Prokudin, P. Sun, and F. Yuan, Phys. Lett. B 750, 533–538 (2015).
- [29]X.-D. Ji, J.-P. Ma, and F. Yuan, Phys. Lett. B 597, 299 (2004).
- [30]J. Collins, L. Gamberg, A. Prokudin, T. Rogers, N. Sato, and B. Wang, Phys. Rev. D 94 (2016).
- [31]A. Bacchetta, F. Delcarro, C. Pisano, M. Radici, and A. Signori, JHEP 2017 (2017).
- [32]A. Bacchetta and A. Prokudin, Nucl. Phys. B 875, 536–551 (2013).
- [33]S. M. Aybat and T. C. Rogers, Phys. Rev. D 83 (2011).
- [34]J. C. Collins and D. E. Soper, Nucl. Phys. B 194, 445 (1982).
- [35]F. Landry, R. Brock, P. M. Nadolsky, and C. P. Yuan, Phys. Rev. D 67, 073016 (2003).
- [36]J.-W. Qiu and X.-F. Zhang, Phys. Rev. Lett. 86, 2724 (2001).
- [37]Z.-B. Kang, B.-W. Xiao, and F. Yuan, Phys. Rev. Lett. 107, 152002 (2011).
- [38]P. Sun, J. Isaacson, C. P. Yuan, and F. Yuan, Int. J. Mod. Phys. A 33, 1841006 (2018).
- [39]A. Prokudin, P. Sun, and F. Yuan, Phys. Lett. B 750, 533 (2015).
- [40]S. Catani, D. de Florian, and M. Grazzini, Nucl. Phys. B 596, 299 (2001).
- [41]H.-L. Lai, M. Guzzi, J. Huston, Z. Li, P. M. Nadolsky, J. Pumplin, and C. P. Yuan, Phys. Rev. D 82, 074024 (2010).
- [42]P. A. Zyla et al. (Particle Data Group), PTEP 2020, 083C01 (2020).