Lecture 2 – Kinematics#
This page works out a selection of the topics and exercises of Lecture 2 by Vincent Mathieu on relativistic kinematics.
Tip
Similar two-to-two kinematics exercises can be found in Question 2.1 of this exercise sheet of the Hadron Spectroscopy Summer School (MTHS).
Experiments and frames#
Particle kinematics is the application of special relativity to elementary particle reactions. In experiments, the “beam” collides onto the “target” with high energy to create other particles.
Each particle has a 4-vector \(p^{\mu}\):
Two frames that are commonly used are the Center of Mass (CoM) frame and the laboratory (Lab) frame.
For a two-particle reaction, \(a+b \to 1 + ... + n\), the frames satisfy the following relations.
Lab frame: \(\vec{p}_b = \vec{0}\)
Target particle \(b\) is at rest, while beam particle \(a\) is moving.CoM frame: \(\vec{p}_a+\vec{p}_b = \vec{0}\)
The total momentum is zero, i.e., the momenta of the two initial state particles have opposite direction.
Masses can be defined as a Lorentz invariant quantity (unaffected by reference frame):
where \(p\) is the 4-vector, \(E\) is the energy, \(\vec{p}\) is the three vector, and \(m\) is the mass.
Two-particle final states#
Let us now consider the case of a two-particle final state (\(n=2\)), i.e., \(a+b \to 1+2\).
When we have a final state with two particles (\(n=2\)), we can define one \(xz\) plane that is spanned by the momenta of the two final state particles in the Lab frame. The CoM frame and the Lab frame can now be related by a boost along the \(z\) axis (the direction of particle \(a\)).
With this definition of the \(xz\) plane, we have:
\(\vec{z}\) is parallel to the beam \(\vec{p}_a\)
\(\vec{y}\) is parallel to \(\vec{p}_a \times \vec{p}_1\)
Lorentz transformations#
Lorentz transformations can be factorized into rotations and boosts along one axis and are required for transforming particle properties, such as position and time, between different inertial frames in relative motion.
Boost along one axis#
When we boost in the direction of the \(z\) axis, only the energy and the \(z\) component of the four-momenta changes, so we can express the Lorentz boost as a 2D matrix:
and its inverse relation,
with
and
Exercise 1
Check that the above boost definition brings the target at rest in the lab frame, that is, that (1) transforms the CoM four-momentum of the target \(b\) into \((E_b^L, p_{b,z}^L) = (m_b, 0)\).
Solution to Exercise 1
We want to show that the boost (1) from the CoM frame to the lab frame brings the target \(b\) to rest, that is,
In the CoM frame, the target moves along the \(z\) axis, so \(p_{b,x}^* = p_{b,y}^* = 0\). A boost along the \(z\) axis does not change these components, so they also vanish in the lab frame and we only have to consider the energy and the \(z\) component.
We first express \(\gamma\) and \(\gamma\beta\) in terms of the CoM energy and momentum of the target. Inserting the velocity (2) into the definition of \(\gamma\) and using the mass-shell relation \(E_b^{*2} - p_{b,z}^{*2} = m_b^2\) gives
which is the result in (3). Substituting these into (1) for particle \(b\), we get
The target therefore has no momentum in the lab frame and its energy equals its mass, so it is at rest.
Rotations#
We use active rotations, meaning that the momentum itself is rotated while the coordinate axes remain fixed.
As an example, take a momentum of unit length in the \(xz\) plane that forms an angle \(\theta\) with the \(z\) axis:
After a rotation over an angle \(\omega\) around the \(y\) axis, the rotated momentum forms an angle \(\theta + \omega\) with the \(z\) axis:
Exercise 2
Check the result of the rotation above.
Solution to Exercise 2
Recall the rotation matrices. Any rotation can be decomposed into rotations around the \(z\) and \(y\) axes:
The rotation matrix around the \(x\) axis is also given for completeness:
With this, we can rewrite the rotation of (4) as
Since \((p_x, p_y, p_z) = (\sin(\theta), 0, \cos(\theta))\), and using the trigonometric identities \(\sin(\theta+\omega) = \sin(\theta)\cos(\omega) + \cos(\theta)\sin(\omega)\) and \(\cos(\theta+\omega) = \cos(\theta)\cos(\omega) - \sin(\theta)\sin(\omega)\), we get:
Example with data#
\(\cos\theta\) in the lab frame#
Exercise 3
The file Two-particles-1.dat contains four-momenta for the reaction \(a+b \to 1+2\). The provided data are measured in the lab frame. Compute and plot the \(\cos\theta_L\) distribution.
Solution to Exercise 3
We first download the data file with gdown and load it with NumPy.
filename = gdown.cached_download(
url="https://indico.ific.uv.es/event/6803/contributions/21220/attachments/11209/15506/Two-particles-1.dat",
path="data/Two-particles-1.dat",
hash="md5:324f71ee88928915f13de5840ac5ede2",
quiet=True,
verify=False,
)
data = np.loadtxt(filename)
data.shape
(15000, 4)
The data file lists the four-momenta \(p_a\), \(p_1\), and \(p_2\) of the reaction \(a+b \to 1+2\) for each event, one after the other. We split them into three arrays.
pa = data[::3].T
p1 = data[1::3].T
p2 = data[2::3].T
pa.shape
(4, 5000)
The four-momentum of the target \(p_b\) follows from four-momentum conservation, \(p_a + p_b = p_1 + p_2\). We check that it is the same for all events.
pb = p1 + p2 - pa
assert pb.std(axis=1).sum().round(5) == 0
pb.mean(axis=1).round(3)
array([ 0.938, 0. , 0. , -0. ])
The mass of particle 1 is given by \(m_1 = \sqrt{E_1^2 - |\vec{p}_1|^2}\):
m1 = np.sqrt(p1[0] ** 2 - p1[1] ** 2 - p1[2] ** 2 - p1[3] ** 2)
m1.mean(), m1.std()
(np.float64(0.13999997797061128), np.float64(9.590363480890062e-07))
The scattering angle \(\theta\) is the angle between the momentum of particle 1 and the beam axis \(z\), so
def p3_length(four_momentum: np.ndarray) -> np.ndarray:
return np.sqrt(np.sum(four_momentum[1:] ** 2, axis=0))
p1_length = p3_length(p1)
cos_theta = p1[3] / p1_length
\(\cos\theta\) in the CoM frame#
Exercise 4
Boost the four-momenta of Exercise 3 to the CoM frame and compute and plot the \(\cos\theta^*\) distribution.
Solution to Exercise 4
We first perform the boost by hand with NumPy.
The CoM frame moves with velocity \(\vec{\beta} = \vec{P}/E\) with respect to the lab frame, where \(P = p_a + p_b = (E, \vec{P})\) is the total four-momentum of the initial state. Since the target is at rest in the lab frame, \(\vec{P} = \vec{p}_a\) and \(E = E_a + m_b\).
P = pa + pb
β_CM = P[1:] / P[0]
β2 = np.sum(β_CM**2, axis=0)
gamma = 1 / np.sqrt(1 - β2)
The energy of particle 1 in the CoM frame follows from the Lorentz boost
with \(\gamma = 1/\sqrt{1-\beta^2}\).
β_dot_p1 = np.sum(β_CM * p1[1:], axis=0)
E1_CM = gamma * (p1[0] - β_dot_p1)
Its three-momentum in the CoM frame is
p1_CM_vec = p1[1:] + ((gamma - 1) / β2 * β_dot_p1 - gamma * p1[0]) * β_CM
p1_CM = np.concatenate((E1_CM[None], p1_CM_vec))
Equation (5) also holds in the CoM frame, so we find \(\cos\theta^*\) with:
p1_CM_length = p3_length(p1_CM)
cos_theta_CM = p1_CM[3] / p1_CM_length
We check the boost with the pylorentz package and compare the \(\cos\theta^*\) values computed in both ways. The method boost_particle() transforms a four-vector from the rest frame of the given momentum into the lab frame. To transform from the lab frame into the CoM frame, we therefore pass it the total four-momentum with reversed three-momentum.
Test passed with relative tolerance 1e-5
Mandelstam variables#
The Mandelstam variables \(s\), \(t\), and \(u\) are Lorentz invariants, defined as
Using the mass-shell relation \(p_i^2 = m_i^2\), their sum only depends on the masses of the particles:
Exercise 5
Check the relation in Equation (6) by using \(p_a+p_b=p_1+p_2\).
Solution to Exercise 5
Expanding the squares in the definitions of \(s\), \(t\), and \(u\) gives
Four-momentum conservation, \(p_b = p_1 + p_2 - p_a\), gives
Substituting this into the expression above, all mixed products cancel and we are left with
where we used the mass-shell relation in the last step.
CoM kinematics#
Every frame-dependent quantity can be expressed in terms of the Mandelstam variables. In the CoM frame, the four-momenta of the reaction \(a+b \to 1+2\) are
so that
and
Exercise 6
Check that \(E_1^* = \frac{s+m^2_1-m_2^2}{2 \sqrt{s}}\) and \(E_2^* = \frac{s+m^2_2-m_1^2}{2 \sqrt{s}}\). Hint: square \(p_2 = (p_a+p_b) - p_1\).
Solution to Exercise 6
Write \(P = p_a + p_b\) for the total four-momentum. In the CoM frame, \(P = (\sqrt{s}, \vec{0})\), so that \(P \cdot p_1 = \sqrt{s}\,E_1^*\). Squaring \(p_2 = P - p_1\) then gives
Solving for \(E_1^*\) gives
Energy conservation in the CoM frame, \(E_1^* + E_2^* = \sqrt{s}\), then gives
Exercise 7
Check that \(|\vec{p}^*_1| = \sqrt{E_1^{*2}-m_1^2} = \frac{\lambda^{1/2}(s,m_1^2,m_2^2)}{2 \sqrt{s}}\), where \(\lambda(x,y,z) = x^2+y^2+z^2 - 2(xy+yz+zx)\) is the Källén function.
Solution to Exercise 7
Inserting \(E_1^*\) from Exercise gives
Expanding the numerator,
so that \(|\vec{p}^*_1| = \lambda^{1/2}(s,m_1^2,m_2^2) / 2\sqrt{s}\). The same steps for the initial state give \(|\vec{p}^*_a| = \lambda^{1/2}(s,m_a^2,m_b^2) / 2\sqrt{s}\).
Since \(-1 \leq \cos\theta^* \leq 1\), Equation (7) shows that \(t\) lies between the values for forward (\(\theta^* = 0\)) and backward (\(\theta^* = \pi\)) scattering,
The forward value \(t_0\) has the smallest magnitude \(|t|\) and is therefore usually denoted \(t_\mathrm{min}\).
With this, the scattering angle in the CoM frame can be expressed in terms of the Mandelstam variables as
Exercise 8
Compute \(\cos\theta^L\), the scattering angle of particle 1 in the lab frame, as a function of \(s\), \(t\), and \(u\).
Solution to Exercise 8
In the lab frame, the target is at rest, \(p_b = (m_b, \vec{0})\). The energies of particles \(a\) and 1 then follow from
and their momenta from the mass-shell relation, in the same way as in Exercise 7:
Just like Equation (7) in the CoM frame, we have \(t = m_a^2 + m_1^2 - 2E_a^L E_1^L + 2|\vec{p}^L_a||\vec{p}^L_1|\cos\theta^L\) in the lab frame. Solving for \(\cos\theta^L\) gives
We check Equations (8) and (9) with the four-momenta of Exercise 3 and compare them with the angles that we computed directly from the four-momenta. The tolerance accounts for the limited number of digits in the data file.
def minkowski_square(p: np.ndarray) -> np.ndarray:
return p[0] ** 2 - np.sum(p[1:] ** 2, axis=0)
def kallen(x: np.ndarray, y: np.ndarray, z: np.ndarray) -> np.ndarray:
return x**2 + y**2 + z**2 - 2 * (x * y + y * z + z * x)
s = minkowski_square(pa + pb)
t = minkowski_square(pa - p1)
u = minkowski_square(pa - p2)
ma_sq, mb_sq, m1_sq, m2_sq = (minkowski_square(p) for p in (pa, pb, p1, p2))
cos_theta_CM_stu = (s * (t - u) + (ma_sq - mb_sq) * (m1_sq - m2_sq)) / np.sqrt(
kallen(s, ma_sq, mb_sq) * kallen(s, m1_sq, m2_sq)
)
cos_theta_lab_stu = (
2 * mb_sq * (t - ma_sq - m1_sq) + (s - ma_sq - mb_sq) * (mb_sq + m1_sq - u)
) / np.sqrt(kallen(s, ma_sq, mb_sq) * kallen(u, mb_sq, m1_sq))
np.testing.assert_allclose(cos_theta_CM_stu, cos_theta_CM, atol=1e-5)
np.testing.assert_allclose(cos_theta_lab_stu, cos_theta, atol=1e-5)
Three-particle final states#
For a reaction \(a+b \to 1+2+3\), we describe the final state with the invariant masses squared of the three pairs of final-state particles, \(s_{ij} = (p_i + p_j)^2\). Only two of them are independent, because their sum is fixed by the invariant mass squared \(M^2 = (p_1+p_2+p_3)^2 = s\) of the final state and the masses of the final-state particles:
Exercise 9
Check the relation in Equation (10).
Solution to Exercise 9
Write \(P = p_1 + p_2 + p_3\), so that \(P^2 = M^2\). Then \(s_{12} = (P - p_3)^2 = M^2 - 2P \cdot p_3 + m_3^2\) and similarly for \(s_{23}\) and \(s_{31}\). Summing the three gives
Dalitz plot analysis#
Lecture 2 comes with two data samples of a reaction with three particles in the final state, Three-particles-1.dat and Three-particles-2.dat. In the following, we identify the reaction and the resonances in these samples with the help of Dalitz plots.
Three-particles-1.dat#
Exercise 10
The file Three-particles-1.dat contains four-momenta for a reaction \(a+b \to 1+2+3\) in the CoM frame. Each event consists of four lines with the four-momenta \((E, p_x, p_y, p_z)\) of particles \(a\), 1, 2, and 3. Which reaction is it?
Solution to Exercise 10
We load the data and split it into the four-momenta of the particles. The four-momentum of the target \(b\) follows from four-momentum conservation.
filename = gdown.cached_download(
url="https://indico.ific.uv.es/event/6803/contributions/21220/attachments/11209/15504/Three-particles-1.dat",
path="data/Three-particles-1.dat",
hash="md5:a49ebfd97ae6a02023291df665ab924c",
quiet=True,
verify=False,
)
data = np.loadtxt(filename)
pa, p1, p2, p3 = (data[i::4].T for i in range(4))
p0 = p1 + p2 + p3
pb = p0 - pa
The four-momentum of the target is opposite to that of the beam, which confirms that the data are given in the CoM frame:
pa.mean(axis=1).round(3), pb.mean(axis=1).round(3)
(array([1.944, 0. , 0. , 1.944]), array([ 2.158, 0. , -0. , -1.944]))
The masses of the particles identify the reaction. Here, \(m_0\) is the invariant mass of the final state, that is, the CoM energy \(\sqrt{s}\).
def mass(p: np.ndarray) -> np.ndarray:
return np.sqrt(mass_squared(p))
def mass_squared(p: np.ndarray) -> np.ndarray:
return p[0] ** 2 - np.sum(p[1:] ** 2, axis=0)
We look up particles with these masses with the particle package. The gluon is excluded, because it is massless like the photon.
from particle import Particle
def find_candidates(
m: float, delta: float = 0.001, charge: float | None = None
) -> list[Particle]:
def identify(p: Particle) -> bool:
if p.pdgid == 21:
return False
if charge is not None and p.charge != charge:
return False
return m - delta < 1e-3 * p.mass < m + delta
return Particle.findall(identify)
initial_state = (
find_candidates(mass(pa).mean(), delta=1e-4)[0],
find_candidates(mass(pb).mean())[0],
)
final_state = tuple(find_candidates(mass(p).mean())[0] for p in (p1, p2, p3))
incoming = ", ".join(p.latex_name for p in initial_state)
outgoing = ", ".join(p.latex_name for p in final_state)
Math(Rf"""
\begin{{aligned}}
\text{{Incoming: }} & {incoming} \\
\text{{Outgoing: }} & {outgoing}
\end{{aligned}}
""")
So this is a photon \(\gamma\) hitting a proton \(p\) at \(\sqrt{s} \approx 4.1\text{ GeV}\), producing an \(\eta\) meson, a pion \(\pi^0\), and a proton \(p\).
Exercise 11
Make a Dalitz plot of the data sample. Which resonances can you identify?
Solution to Exercise 11
A Dalitz plot shows two of the three invariant masses squared \(s_{ij}\) against each other. Equation (10) shows that the third one carries no additional information. We check this relation with the data:
np.testing.assert_allclose(
mass_squared(p1 + p2) + mass_squared(p2 + p3) + mass_squared(p3 + p1),
mass_squared(p0) + mass_squared(p1) + mass_squared(p2) + mass_squared(p3),
rtol=1e-5,
)
A resonance in the \(ij\) subsystem shows up as a band of constant \(s_{ij}\) in the Dalitz plot. We estimate the positions \(R_{ij}\) of these bands from the peaks in the distributions of \(s_{12}\), \(s_{23}\), and \(s_{31}\).
s = {
"12": mass_squared(p1 + p2),
"23": mass_squared(p2 + p3),
"31": mass_squared(p3 + p1),
}
peaks = {ij: find_peak(s_ij) for ij, s_ij in s.items()}
{ij: round(float(np.sqrt(r)), 3) for ij, r in peaks.items()}
{'12': 1.318, '23': 1.235, '31': 1.564}
The particle candidates for \(R_{12} \to \eta\pi^0\) and \(R_{23} \to \pi^0 p\) are:
find_candidates(np.sqrt(peaks["12"]), delta=0.01, charge=0)
[<Particle: name="a(2)(1320)0", pdgid=115, mass=1318.2 ± 0.6 MeV>,
<Particle: name="Xi0", pdgid=3322, mass=1314.86 ± 0.20 MeV>,
<Particle: name="Xi~0", pdgid=-3322, mass=1314.86 ± 0.20 MeV>]
find_candidates(np.sqrt(peaks["23"]), delta=0.01, charge=+1)
[<Particle: name="Delta(1232)~+", pdgid=-1114, mass=1232.0 ± 2.0 MeV>,
<Particle: name="Delta(1232)+", pdgid=2214, mass=1232.0 ± 2.0 MeV>,
<Particle: name="b(1)(1235)+", pdgid=10213, mass=1230 ± 3 MeV>,
<Particle: name="a(1)(1260)+", pdgid=20213, mass=1230 ± 40 MeV>]
The \(\Xi^0\) baryon cannot decay into two mesons, so the \(\eta\pi^0\) resonance is the \(a_2(1320)\). Likewise, the \(b_1(1235)^+\) and \(a_1(1260)^+\) mesons and the anti-baryon \(\bar\Delta(1232)^+\) cannot decay into \(\pi^0 p\), so the \(\pi^0 p\) resonance is the \(\Delta(1232)^+\).
The \(p\eta\) band is wider. Since the \(\eta\) has isospin 0, a \(p\eta\) resonance has isospin 1/2, which excludes the \(\Delta\) states. The remaining candidates within 50 MeV are:
[
p
for p in find_candidates(np.sqrt(peaks["31"]), delta=0.05, charge=+1)
if p.pdgid > 0 and np.isclose(p.I, 1 / 2)
]
[<Particle: name="N(1520)+", pdgid=2124, mass=1515 ± 5 MeV>,
<Particle: name="N(1535)+", pdgid=22212, mass=1530 ± 15 MeV>]
Of these, the \(N(1535)^+\) is known for its strong coupling to \(N\eta\), so it is the most likely candidate, even though the peak lies somewhat above its nominal mass. In summary, the reaction is \(\gamma p \to \eta\pi^0 p\) with the resonances \(a_2(1320) \to \eta\pi^0\), \(\Delta(1232)^+ \to \pi^0 p\), and \(N(1535)^+ \to p\eta\).
Exercise 12
What can you say about the spin of the resonances?
Solution to Exercise 12
A resonance with spin \(J\) generally produces an angular distribution of its decay products in its rest frame. Along a resonance band in the Dalitz plot, the other invariant masses are linear functions of the cosine of this decay angle, so an angular distribution shows up as a modulation of the event density along the band. We therefore plot the distribution of events within each band.
Apart from the peaks at the edges, where the bands cross the other resonance bands, the event density along each band is roughly flat. These data therefore show no sign of the spin of the resonances, even though the \(a_2(1320)\) has \(J=2\) and the \(\Delta(1232)\) has \(J=3/2\): the sample was apparently generated without an angular dependence.
Three-particles-2.dat#
Exercise 13
Answer the same questions for Three-particles-2.dat.
Solution to Exercise 13
filename = gdown.cached_download(
url="https://indico.ific.uv.es/event/6803/contributions/21220/attachments/11209/15511/Three-particles-2.dat",
path="data/Three-particles-2.dat",
hash="md5:831aee9fd925c43e8630edc6783ab28d",
quiet=True,
verify=False,
)
data = np.loadtxt(filename)
pa, p1, p2, p3 = (data[i::4].T for i in range(4))
p0 = p1 + p2 + p3
pb = p0 - pa
initial_state = (
find_candidates(mass(pa).mean(), delta=1e-4)[0],
find_candidates(mass(pb).mean())[0],
)
final_state = tuple(find_candidates(mass(p).mean())[0] for p in (p1, p2, p3))
incoming = ", ".join(p.latex_name for p in initial_state)
outgoing = ", ".join(p.latex_name for p in final_state)
Math(Rf"""
\begin{{aligned}}
\text{{Incoming: }} & {incoming} \\
\text{{Outgoing: }} & {outgoing}
\end{{aligned}}
""")
This is again \(\gamma p \to \eta\pi^0 p\), but at a lower CoM energy of \(\sqrt{s} \approx 3.35\text{ GeV}\).
s = {
"12": mass_squared(p1 + p2),
"23": mass_squared(p2 + p3),
"31": mass_squared(p3 + p1),
}
peaks = {ij: find_peak(s_ij) for ij, s_ij in s.items()}
{ij: round(float(np.sqrt(r)), 3) for ij, r in peaks.items()}
{'12': 1.317, '23': 1.239, '31': 1.585}
The bands lie at the same positions as in Three-particles-1.dat, so the same resonances \(a_2(1320)\), \(\Delta(1232)^+\), and \(N(1535)^+\) contribute. Because of the lower energy, the Dalitz plot is smaller and the bands overlap more. Also in this sample, the event density along the bands shows no modulation beyond the crossings with the other bands.