MathIsimple

Probability Theory – Problem 20: Determine whether the simple random walk on is recurrent or transient, and prove your conclusion

Question

Consider an infinite connected graph G=(V,E)G=(V,E) with uniformly bounded vertex degrees. Define G3=(V3,E)G^{3}=(V^{3},\mathcal{E}), where ((v1,v2,v3),(u1,u2,u3))E\left(\left(v_{1},v_{2},v_{3}\right),\left(u_{1},u_{2},u_{3}\right)\right)\in\mathcal{E} if and only if u1,u2,u3,v1,v2,v3Vu_{1},u_{2},u_{3},v_{1},v_{2},v_{3}\in V and there exists i{1,2,3}i\in\{1,2,3\} such that

uj=vj for ji,(ui,vi)E.u_j = v_j \text{ for } j \ne i,\quad (u_i,v_i)\in E.

Determine whether the simple random walk on G3G^{3} is recurrent or transient, and prove your conclusion.

Step-by-step solution

Step 1. Consider an infinite connected graph G=(V,E)G=(V,E) with uniformly bounded vertex degrees; write M=supvVdeg(v)<M = \sup_{v \in V} \deg(v) < \infty. The graph G3G^3 is the threefold Cartesian product of GG. In G3G^3, every vertex v=(v1,v2,v3)\mathbf{v} = (v_1, v_2, v_3) has degree degG3(v)=degG(v1)+degG(v2)+degG(v3)\deg_{G^3}(\mathbf{v}) = \deg_G(v_1) + \deg_G(v_2) + \deg_G(v_3). Since the degrees in GG are uniformly bounded, degG3(v)3M\deg_{G^3}(\mathbf{v}) \le 3M. Recurrence of the simple random walk on a graph is intimately connected with the electrical resistance network properties of the graph. By the Nash-Williams criterion or Rayleigh's monotonicity principle, if one can contract or embed the graph into a structure of known recurrence or transience type, this aids the determination. Step 2. For the dd-dimensional integer lattice Zd\mathbb{Z}^d, the simple random walk is transient when d3d \ge 3. Since GG is an infinite connected graph, it necessarily contains an infinite simple path or some other infinite structure. Because GG is infinite and connected, starting from any vertex there exists at least one infinite self-avoiding path P={x0,x1,x2,}VP = \{x_0, x_1, x_2, \dots\} \subseteq V with (xk,xk+1)E(x_k, x_{k+1}) \in E. Step 3. Consider the subgraph H=P3G3H = P^3 \subseteq G^3. The path PP is isomorphic to the one-dimensional integer lattice Z\mathbb{Z} (or its half-line N\mathbb{N}). Therefore the structure of HH is isomorphic to (a portion of) the three-dimensional lattice Z3\mathbb{Z}^3 (or N3\mathbb{N}^3). For the electrical resistance network on G3G^3, assign resistance 11 to every edge. By Rayleigh's monotonicity principle, removing edges from a graph (i.e., increasing resistances) can only increase the effective resistance. If a subgraph is transient, then the original graph is also transient. Step 4. Specifically, examine the subgraph HH of G3G^3 generated by the path PP. The vertex set of HH is VP=P×P×PV_P = P \times P \times P, and its edge set consists only of edges corresponding to adjacent vertices in PP. In fact, HH is isomorphic to the three-dimensional lattice graph N3\mathbb{N}^3 (if PP is one-sided) or Z3\mathbb{Z}^3 (if PP is two-sided). It is well known that the simple random walk on the three-dimensional lattice Z3\mathbb{Z}^3 is transient. The same holds for the half-space or octant version N3\mathbb{N}^3. Step 5. Since HH is a subgraph of G3G^3 and the simple random walk on HH is transient, the effective resistance from the origin to infinity is finite. By the comparison theorem, the additional edges in G3G^3 only decrease the effective resistance, thereby increasing the escape probability from the origin to infinity. Therefore the simple random walk on G3G^3 is also transient.

Final answer

Transient. QED.

Marking scheme

Based on the official solution provided, the following is the grading rubric for this problem.


1. Checkpoints (max 7 pts total)

Score exactly one chain | take the maximum subtotal among chains; do not add points across chains.

Chain A: Subgraph Embedding / Rayleigh Monotonicity Principle (Official Solution Idea)

  • [Step 1] State or argue: since GG is an infinite connected graph, GG contains an infinite simple path (or self-avoiding path) PP, and PP is isomorphic to N\mathbb{N} (or Z\mathbb{Z}). [1 pt]
  • *Note: If the student merely mentions that GG has infinitely many vertices without addressing the path structure, no credit is awarded.*
  • [Step 2] Construct or identify the subgraph H=P3H = P^3 (i.e., P×P×PP \times P \times P) in G3G^3, and observe that this subgraph is isomorphic to the three-dimensional lattice graph N3\mathbb{N}^3 (or Z3\mathbb{Z}^3). [2 pts]
  • *Note: If the student identifies the subgraph P3P^3 but does not explicitly connect it to the structure of N3\mathbb{N}^3 or Z3\mathbb{Z}^3, award 1 pt.*
  • [Step 3] Cite the known result: the simple random walk on the three-dimensional lattice N3\mathbb{N}^3 (or Z3\mathbb{Z}^3) is transient. [1 pt]
  • [Step 4] Apply Rayleigh's Monotonicity Principle or the resistance network comparison theorem to argue: [2 pts]
  • Observe that G3G^3 is a supergraph of HH, i.e., G3G^3 has more edges than HH (equivalently, smaller effective resistance). (1 pt)
  • Argue that "a graph containing a transient subgraph is itself transient" (i.e., Reff(G3)Reff(H)<R_{\text{eff}}(G^3) \le R_{\text{eff}}(H) < \infty). (1 pt)
  • [Step 5] State the final conclusion: the simple random walk on G3G^3 is transient. [1 pt]
  • *Note: The conclusion must be supported by the preceding argument; if the reasoning contains logical errors, this point is not awarded.*

Total (max 7)


2. Zero-credit items

  • Stating only the conclusion "transient" without any valid mathematical argument or justification.
  • Stating only "because the dimension is 3" without defining the dimension of the graph or proving the specific relationship between volume growth rate and recurrence (this is a heuristic observation, not a rigorous proof).
  • Proving the result only for the special case G=ZG=\mathbb{Z} without generalizing to the general case or invoking the subgraph comparison principle.
  • Merely copying the problem definitions (such as the definition of G3G^3) or listing known conditions (such as "uniformly bounded degrees").

3. Deductions

  • −1 (Logic Gap): Reversing the direction in the comparison principle (e.g., erroneously claiming "since the subgraph is recurrent, the supergraph is recurrent," or confusing the relationship between resistance magnitude and recurrence).
  • −1 (Conclusion): Confusing terminology by writing "recurrent" instead of "transient," even though the derivation implies finite resistance.
  • −2 (Major Concept): Attempting to use a criterion designed for establishing recurrence (such as Nash-Williams' 1Bn=\sum \frac{1}{|\partial B_n|} = \infty) to force a proof of transience, resulting in a logical contradiction.
  • Cap at 3/7: Attempting a volume growth argument claiming V(n)cn3V(n) \ge cn^3 implies transience, without citing a specific theorem applicable to general bounded-degree graphs (note: volume growth alone is insufficient for general graphs; typically an isoperimetric inequality is also needed, whereas the lattice embedding method is the most direct and rigorous approach).
Ask AI ✨