What makes an estimate uniform?

Suppose every member of a sequence of regulated theories has a positive spectral gap. Does the limiting theory have one? Not necessarily. The gaps may shrink to zero. A uniform estimate is an estimate whose relevant constants remain controlled throughout the limit being taken.

The phrase is incomplete until we specify which parameter varies, which units are used, and what remains fixed.

A disappearing gap

Prerequisites: exponential correlators and lattice/physical units.
Outcomes: identify which constants depend on a regulator; pass a genuinely uniform bound to a limit; diagnose a positive gap that disappears.

Take a fixed Hilbert space with an orthonormal vacuum $\Omega$ and excited vectors $e_1,e_2,\ldots$. Define

$$H_n\Omega=0,\qquad H_ne_k=\left(\frac1n+\frac1k\right)e_k.$$

For each $n$, all excited energies lie above $1/n$, and their infimum is $\Delta_n=1/n$. The limiting operator has excited energies $1,1/2,1/3,\ldots$, arbitrarily close to zero. Its vacuum remains unique, but it is gapless. Positivity of every $\Delta_n$ did not provide a positive lower bound independent of $n$.

For a correlator, an estimate of the form

$$|C_L(t)|\leq A_L e^{-m_Lt}$$

must therefore be read with its constants attached. Does $m_L$ stay positive as the box size $L$ grows? Is $A_L$ controlled for the observables being compared? Is convergence of $C_L$ itself established?

A limit that can be passed

Suppose, in a common physical normalization, that

$$|C_n(t)|\leq A e^{-mt}$$

holds for every $n$ and $t\geq0$, with fixed finite $A$ and fixed $m>0$. If $C_n(t)\to C(t)$ for each $t$, then

$$|C(t)|\leq A e^{-mt}.$$

This elementary argument shows the value of uniformity. It does not by itself establish that the limiting correlators define the desired QFT or that this single correlation bound detects its whole spectrum.

Physical units change the appearance

Let the lattice spacing be $a$, and let $n$ denote a temporal separation in lattice steps. Physical time is $t=an$. A physical exponential reads

$$e^{-m_{\mathrm{phys}}t} =e^{-(a m_{\mathrm{phys}})n}.$$

The dimensionless lattice mass is $m_{\mathrm{latt}}=a m_{\mathrm{phys}}$. Along a continuum sequence with a fixed positive physical mass, $m_{\mathrm{latt}}$ tends to zero. Demanding a fixed positive lower bound on $m_{\mathrm{latt}}$ would therefore ask for the wrong scaling behavior.

Conversely, knowing $m_{\mathrm{latt}}>0$ at every spacing gives no guarantee that $m_{\mathrm{latt}}/a$ tends to a finite positive limit. The reference scale and trajectory matter.

Different limits need different controls

Common parameters include lattice spacing $a$, physical box size $L$, rank $N$, compactification radius, and truncation size. A theorem uniform in $L$ at fixed $a$ need not be uniform in $a$. A numerical bootstrap bound valid at each truncation need not already establish convergence of the hierarchy.

When reading a result, write a small “constant ledger”: what each constant depends on, which variables approach limits, and whether the hypotheses remain valid along that path. This is often more revealing than the abstract.

Exercise. Compare $m_{\mathrm{latt}}(a)=a$, $a^2$, and $\sqrt a$ for $a\to0$, with the same declared physical units.

Solution. The corresponding physical masses are $1$, $a$, and $1/\sqrt a$. They approach a finite positive value, zero, and infinity respectively. All three lattice masses are positive and tend to zero.

Continue

Use weak convergence and tightness to distinguish a subsequential law from a unique identified theory. Limits and uniformity puts the constant ledger alongside the other construction requirements.

Try the continuum lab and read the continuum-limit lesson. Section 6 of Jaffe and Witten explains the importance of uniform estimates in constructive approaches. For an example with precise coupling and volume hypotheses, see Shen, Zhu, and Zhu.