JERRY E. SMITH, Circuit Judge:

James Brown challenges his convictions on the ground that the government violated his right to due process by withholding materially favorable evidence that it possessed pre-trial. See Brady v. Maryland, 373 U.S. 83, 83 S.Ct. 1194, 10 L.Ed.2d 215 (1963). Because the district court did not clearly err in holding that the evidence was not material, we affirm.

I.

This appeal arises from an earlier trial relating to the Enron scandal. See United States v. Brown (Brown I), 459 F.3d 509, 513 (5th Cir.2006). At years' end 1999, Merrill Lynch purchased an equity interest in three barge-mounted power generators off the Nigerian coast from Enron Corporation ("Enron") for $28 million, with Merrill Lynch paying Enron $7 million and Enron loaning Merrill Lynch the balance. Enron booked a roughly $12 million profit on the transaction. The government contended that the sale was a sham whose sole purpose was to allow Enron artificially to enhance its fourth-quarter earnings to meet forecasts. According to the government, the transaction was not a true sale, because Enron did not actually sell a stake in the barges but instead secretly promised that a company run by Andrew Fastow, Enron's CFO, would buy back the stake in the barges from Merrill Lynch within six months for a guaranteed 15% return plus a $250,000 "advisory fee." In other words, the government alleged Enron just loaned out the stake in the barges to Merrill Lynch, risk-free and with a guaranteed return, but made it seem like a sale so that it could book a pretend profit.

Brown was a managing director at Merrill Lynch and the head of its Strategic Asset and Lease Finance group at the time of the transaction. He testified to a grand jury that, to his knowledge, Enron had never promised that it would buy back Merrill Lynch's equity in the barges within six months of the purported sale.

The government indicted Brown, charging him with, as relevant here, perjury and obstruction of justice, alleging that Enron executives orally guaranteed to repurchase Merrill Lynch's equity stake in the barges, and Brown knowingly lied to the grand jury about his understanding of the transaction.1 Specifically, the indictment quoted the following testimony and alleged that the underlined portions were false:

Also relevant is the following testimony elaborating on Brown's understanding of the transaction:

We summarize the detailed evidence presented at trial relating to the perjury and obstruction-of-justice charges: On December 22, 1999, Merrill Lynch employee Tina Trinkle participated in a conference call (the "Trinkle call") that included Brown. Trinkle testified that, during the call, "[s]omebody at Enron" promised Merrill Lynch that the Nigerian barges would be bought back, and a Merrill Lynch executive (possibly Brown himself; Trinkle was not sure) rejected putting that guarantee in writing, because it would not allow "the right accounting treatment." Merrill Lynch employees asserted during the call that someone at Enron—they did not say who—had given them "his word" and "his strongest verbal assurances" of a buyback. No lawyers participated in the call.

Trinkle said Brown "was very negative on the deal, and he felt that it had a lot of risks."2 For example, Trinkle said Brown was concerned about the "political risk" involved in the transaction (because the barges were in Nigeria). Brown's notes also indicate that he was concerned about the "reputational risk" of "aid[ing]/abet[ting] Enron income stmt. manipulation," and he communicated those concerns to Bill Fuhs, a vice-president working under him.

Katherine Zrike, chief counsel for Merrill Lynch's investment banking division, said Bob Furst, a managing director at Merrill Lynch and the investment banker responsible for the Enron account, told her, before the Trinkle call, that "the only agreement between Enron and Merrill Lynch was that Enron would help Merrill Lynch re-market the barges," that is, do its best to find a third party to purchase them from Merrill Lynch. Indeed, a memorandum dated the day before the Trinkle call and sent from Furst to Brown said, "Enron is viewing this transaction as a bridge to permanent equity and they believe our hold will be for less than six months." (Emphasis added.)

After the Trinkle call, that same day, Zrike convened a meeting of Merrill Lynch's Debt Markets Commitment Committee ("DMCC"), in which Brown participated, at which "everybody was agreeing" that there could not be a buyback of Merrill Lynch's equity interest in the barges, because that would not permit Enron legally to account for the transfer of the barges to Merrill Lynch as a sale. Furst stated at the meeting that the "`real agreement with Enron is only to re-market.'" The DMCC did not approve the transaction but instead opted to have Dan Bayly, head of investment banking at Merrill Lynch, and his boss, Tom Davis, review it for approval or rejection.

Shortly thereafter, Zrike, Bayly, and others (but not Brown) met with Davis in Davis' conference room, where the deal was explained to Davis. Zrike said they "talked about the fact that this needed to be a true sale and, therefore, all risks of loss and all risks associated with owning the barge would pass to Merrill Lynch for the time that it owned the barges." Zrike mentioned the risks of dealing with a property located in Nigeria, and there was a discussion about the fact that there had been no due diligence on the barges. Davis ultimately approved the deal, although he was "not happy" about it.

Brown went on vacation the day after the Trinkle call and DMCC meeting.3 That day, Fastow conducted a conference call with Merrill Lynch that did not include Brown or Merrill Lynch's chief counsel, Zrike. No one who participated in the call testified about its contents, but Eric Boyt, an in-house accountant at Enron, testified that Daniel Boyle, an Enron finance executive who participated in the call, told him right after the call that Fastow had guaranteed a buyback with 15% return in six months if a buyer could not be found. Ben Glisan and Michael Kopper, both high-ranking Enron finance executives, also testified that Fastow and Enron Treasurer Jeff McMahon later told them that they had "promised" Merrill Lynch that they would make sure it was out of the Nigerian barge transaction within six months.

There are also contemporaneous emails from Glisan and James Hughes, another Enron executive, saying, respectively, that, "[t]o be clear, Enron is obligated to get Merrill Lynch out of the deal [by] June 30" and that if "no one will take the Merrill Lynch position, then Enron will inherit it." Finally, there is an unsigned, undated internal Merrill Lynch document from sometime before December 31, 1999, that says that Enron "assured" Merrill Lynch that it "will be taken out of our investment within six months."

The engagement letter itself, which was signed by Brown, makes no mention of a buyback guarantee or a remarketing agreement.4 An earlier draft of the letter, written by an associate in Brown's department and sent to Fuhs on December 23, 1999 (while Brown was away), says, however, that Merrill Lynch's stake in the barges "will be subsequently sold to third party investors or purchased by Enron or an affiliate" and that Merrill Lynch would receive a 15% annualized return on its investment. Enron executive Boyle struck that language before the final draft.

Sean Long, head of the Enron group that oversaw the Nigerian barge project in Africa, testified that no one at Merrill Lynch "contact[ed] [him] at all with respect to the barges" between January and June 2000; that is, Merrill Lynch did not follow-up on the barges after it bought them, which indicates that it knew they would be bought back. Long also testified that Boyle had told him "that a senior person at Enron gave assurances to a senior person at Merrill Lynch that they would not get hurt by the transaction."

In June 2000, six months after Merrill Lynch obtained its interest in the barges, LJM25—a partnership that, according to Kopper, was "set up by ... Fastow, to raise private equity for deals that were to be done with Enron"—purchased Merrill Lynch's equity interest in the barges at a 15% annualized return. Fastow was LJM2's general partner. Kopper testified that "Enron would use LJM as essentially an off-ramp on deals that they needed to use to make earnings for any given quarter." Enron would "warehouse" assets in LJM2 for six months to "misstate" that it had sold them. Kopper referred to "this Nigerian barge deal" as a transaction involving such a misstatement, and

Furthermore, an Enron document, the "Benefits to Enron Summary," dated June 29, 2000, states that "Enron sold barges to Merrill Lynch (ML) in December of 1999, promising that Merrill would be taken out by sale to another investor by June, 2000." (Emphasis added).

A couple of emails more directly implicate Brown. After LJM2's purchase of the interest in the barges, Fuhs had an email exchange with Brown in which Fuhs said, "Enjoy the barges on the other side of this trade and good luck." Fuhs was referring to the fact that Brown had an investment in LJM2, which now had a stake in the barges. Brown responded, "thanks bill ... wanna buy a barge?" to which Fuhs replied, "only if I can have a guaranty [sic] of make-whole at par + return in case of civil unrest/war." (Emphasis added).

More significantly, Brown sent an email in March 2001 about an unrelated transaction, saying he would "support an unsecured deal provided we had total verbal assurances from [the company's CEO or CFO]," explaining that "[w]e had a similar precedent with Enron last year, and we had Fastow get on the phone with Bayly and lawyers and promise to pay us back no matter what. Deal was approved and all went well." (Emphasis added).6

In short, there is considerable evidence that Enron executives orally promised Merrill Lynch that it or a third party would buy back the barges within six months. An email sent by Brown plainly shows his awareness of that promise. But one can perhaps question, as Brown's attorney did at closing argument and during Brown's original appeal, whether Enron executives really made a "promise" as one might understand it in the commercial context—namely, a binding commitment—or whether it merely meant giving one's not-always-reliable word, what Brown referred to in his grand jury testimony as "strong comfort."7

II.

The jury convicted Brown of perjury and obstruction of justice. A divided panel affirmed, with Judge DeMoss dissenting on the ground that the evidence was insufficient for a reasonable jury to find Brown had lied, because Fastow's "promise" was not a legally enforceable commitment and thus was not a true promise.8

III.

Brown now challenges his convictions on the ground that the government violated his right to due process by withholding materially favorable evidence that it possessed pre-trial. Brown focuses on three allegedly new pieces of evidence: (1) The FBI's notes of its interview with Fastow, (2) Senate investigators' notes of their interview with McMahon, and (3) transcripts of Zrike's pretrial testimony before the grand jury and the SEC.

The government disclosed pre-trial two letters that it says fairly summarized the exculpatory aspects of the Fastow and McMahon notes and the Zrike testimony. The government also showed the McMahon notes and the Zrike testimony to the district court in camera before Brown's trial, and the court did not find it necessary for the government to produce anything more than the summary letters. The government concedes that it did not submit the Fastow notes to the district court for in camera review.

Brown argues that there are significant differences between the Fastow and McMahon raw notes and the Zrike transcript, on the one hand, and the government letters purportedly summarizing them, on the other hand. The district court decided that the government did not violate its Brady obligation, holding that the government did not suppress favorable evidence and that, even if it did, it was not material.

IV.

To establish a Brady violation, the defendant must prove that (1) the prosecution suppressed evidence, (2) it was favorable to the defendant, and (3) it was material.9 The good or bad faith of the prosecution in suppressing evidence is irrelevant. Kyles v. Whitley, 514 U.S. 419, 432, 115 S.Ct. 1555, 131 L.Ed.2d 490 (1995) (citing Brady, 373 U.S. at 87, 83 S.Ct. 1194). But evidence is not suppressed "`if the defendant knows or should know of the essential facts that would enable him to take advantage of it.'" Skilling, 554 F.3d at 575 (quoting United States v. Runyan, 290 F.3d 223, 246 (5th Cir.2002)). To have been suppressed, the evidence must not have been discoverable through the defendant's due diligence.10

Evidence is material if there is "`a reasonable probability that, had the evidence been disclosed to the defense, the result of the proceeding would have been different.'" United States v. Bagley, 473 U.S. 667, 682, 105 S.Ct. 3375, 87 L.Ed.2d 481 (1985) (citing Strickland v. Washington, 466 U.S. 668, 694, 104 S.Ct. 2052, 80 L.Ed.2d 674 (1984)). In other words, "[t]he question is not whether the defendant would more likely than not have received a different verdict with the evidence, but whether in its absence he received a fair trial, understood as a trial resulting in a verdict worthy of confidence." Kyles, 514 U.S. at 434, 115 S.Ct. 1555. A "reasonable probability" exists when the government's suppression of evidence "`undermines confidence in the outcome of the trial.'" Id. (quoting Bagley, 473 U.S. at 678, 105 S.Ct. 3375). To prove a reasonable probability of a different result, the "likelihood of a different result must be substantial, not just conceivable." Harrington v. Richter, ___ U.S. ___, 131 S.Ct. 770, 792, 178 L.Ed.2d 624 (2011) (citing Washington, 466 U.S. at 693, 104 S.Ct. 2052). A "reasonable probability" is less than "`more likely than not,'" but the difference "is slight and matters `only in the rarest case.'" Id. (quoting Washington, 466 U.S. at 693, 697, 104 S.Ct. 2052).11

There is no difference between exculpatory and impeachment evidence for purposes of Brady. Kyles, 514 U.S. at 433, 115 S.Ct. 1555 (citing Bagley, 473 U.S. 667, 105 S.Ct. 3375). The suppressed evidence need not be admissible to be material under Brady; but it must, somehow, create a reasonable probability that the result of the proceeding would be different.12 We assess the materiality of the suppressed evidence cumulatively, not item by item.13 Once a Brady violation has been shown, there is no need for further harmless-error review, id. at 435, 115 S.Ct. 1555, and a new trial is the prescribed remedy, not a matter of discretion.14

A.

We generally review whether the government violated Brady de novo, Skilling, 554 F.3d at 578, although even when reviewing a Brady claim de novo, "we must proceed with deference to the factual findings underlying the district court's decision," United States v. Sipe, 388 F.3d 471, 479 (5th Cir.2004). But we have an exception to our general rule of de novo review: Where, as is partially the case here, "a district court has reviewed potential Brady material in camera and ruled that the material was not discoverable, we review [that] decision only for clear error."15 The district court's finding is clearly erroneous if, on the entire evidence, we are left with a "definite and firm conviction" that a mistake has been committed. United States v. U.S. Gypsum Co., 333 U.S. 364, 395, 68 S.Ct. 525, 92 L.Ed. 746 (1948).

Thus, with respect to suppression and favorability—the first two prongs of the Brady test—we apply two different standards of review: Because the Fastow notes were never seen by the district court before trial, we review whether they are discoverable de novo (with deference to the district court's underlying factual findings). But because the court did review the McMahon notes and Zrike testimony pre-trial, we review its decision as to those items for clear error. And because we conclude that the withheld portions of the Fastow notes are not favorable to Brown, all favorable evidence was reviewed by the court in camera pre-trial. We therefore review materiality for clear error as well.16

B.

The first potential Brady item is the FBI's notes from its interview with Fastow, which were never disclosed to Brown, although the government did disclose a letter summarizing the notes. The issue is whether any evidence favorable to Brown in the Fastow notes was suppressed, in light of the government's disclosure letter.

Brown argues that the FBI's raw notes, unlike the government's disclosure letter, referenced a "best efforts" agreement, said that Fastow "never used the word promise," and contained assorted other, similar statements, such as "summary not consistent w/ [Fastow]'s memory b/c not word `promise.'" The district court held that no favorable information from the notes was suppressed, because the disclosure letter did reveal that Fastow said that "Enron was the marketing agent, but could not make anyone buy at a specified time, price or return" and that "Fastow deliberately avoided the word `guarantee' and knew that he could not give a verbal or written guarantee on the deal without jeopardizing the accounting treatment Enron needed."

We agree with the district court. Saying that Enron "could not make anyone buy" or that Fastow "deliberately avoided the word `guarantee,'" knowing that he "could not give a verbal or written guarantee," conveys essentially the same information as "never used the word promise" or "obligation to use `best efforts.'" Moreover, any potential exculpatory value of the passages from the Fastow notes that were not disclosed to the defense is eliminated when we read them in context rather than looking just to the portions of the sentences that Brown cherry-picks.

The notes say, to give only a few examples, (1) "It was [Enron's] obligation to use `best efforts' to find 3rd party takeout + went on to say there would be 3rd party b/c AF is manager of third party," (emphasis added); (2) "LJM was 3rd party + was already found;" (3) "[Fastow] told [Merrill Lynch] that [Enron] would get [Merrill Lynch] out, would get [illegible] or LJM to buy out;" and (4) "Come June 2000, if [Enron] did not have a buyer then LJM would step in to buy out." Thus, the sentences that Brown cites from the Fastow notes do not say that the agreement as a whole was a "best efforts" agreement, pace Brown's testimony; they say only that Enron would use its "best efforts" to find a buyer but that Fastow guaranteed that LJM2, which he controlled, would be that buyer if no one else was found. Indeed, Fastow admitted that, "[i]f call was transcribed—it should have blown the accounting."

That is how this court interpreted the same statements in Fastow's notes in Skilling17 in rejecting an essentially identical Brady claim.18 The relevant passages, read in full, thus corroborate the government's position, not Brown's, by showing that Fastow did promise a buyback by LJM2. Thus, the government's disclosure letter accurately stated that "Fastow did not say Enron would buy back the barges, but represented instead that a third party would," and no favorable evidence was suppressed.

Second, Brown highlights a portion of the notes that says,

The district court noted that those statements were "arguably ... suppressed" but decided they were not material. The information indicating that Fastow used different terminology with his employees than he did with Merrill Lynch was omitted from the government's disclosure letter, however, and was not otherwise available to Brown. So it was suppressed.

But it was not favorable to Brown. Read in context, Fastow's statements say only that Fastow was hiding LJM's role in the barges transaction from his subordinates, not that there was no promise. Fastow's promise to Merrill Lynch, as reflected in the notes, was that LJM would buy back the interest in the barges if a third-party buyer could not be found. Skilling, 554 F.3d at 589. Indeed, immediately preceding the passage that Brown cites, Fastow explained, "By referencing [that he was LJM's] General Partner [in the call with Merrill Lynch], was in effect giving the guarantee.... [I]f LJM not buyer then [Enron] will take necessary steps to make sure [Merrill Lynch] not owner."19

Fastow then goes on to say, in the passage Brown cites, that he told subordinates that Enron would buy back the interest in the barges, because if he told them about LJM, they would lose motivation to find a third-party buyer. That is the only possible explanation for his statement, "Internally said Enron would buy back. Unit less motivated if knew of LJM." (Emphasis added.) That Fastow told his subordinates that Enron would buy back so that they did not know LJM would do so supports, rather than undermines, the government's argument that Fastow made a promise that LJM would buy. Indeed, we so held in Skilling, explicitly rejecting the notion that this portion of the notes implied that Fastow admitted to lying to subordinates that there was a promise.20

Brown's argument thus boils down to the proposition that we should consider the passages he cites to be exculpatory because he could have put some misleading spin on them to the jury. But because the only fair reading of those passages is an inculpatory one, the government is correct that no favorable evidence was suppressed.

C.

Brown claims the government withheld exculpatory portions of (1) the Senate Permanent Subcommittee on Investigations's notes from its interview with McMahon and (2) Zrike's grand jury and SEC testimony. Favorable information was plainly suppressed from McMahon's notes, and we will assume arguendo that favorable information from Zrike's testimony was suppressed as well. Nevertheless, the district court did not clearly err in holding that the suppressed information was not cumulatively material.21

The McMahon notes contain numerous passages that unequivocally state that it was McMahon's understanding that there was only a "best efforts" agreement and no "promise," whereas the government's disclosure letter says only that McMahon "does not recall" a guaranteed buyback. The district court thus clearly erred in holding that the government's disclosure letter fully disclosed the contents of the notes: "No" is not the same thing as "I do not recall." But despite the exculpatory nature of the suppressed portions of the McMahon notes, Brown could have made only very little use of them.

The parties stipulated that McMahon was unavailable as a witness because he would invoke his Fifth Amendment privileges if called to testify, so access to the McMahon notes would not have aided Brown in that sense. At most, Brown could have used McMahon's statements from the Senate subcommittee notes to impeach Glisan's and Kopper's testimony that McMahon told them there was a buyback "promise."22 But McMahon's statements to Glisan and Kopper were merely cumulative evidence: Glisan and Kopper also gave unimpeached testimony that Fastow told them he promised Merrill Lynch that he would buy the barges back; Trinkle, Boyt, and Long all testified to the same effect; and multiple Enron and Merrill Lynch documents, including Brown's email, said there was a promise.

The "impeached testimony of a witness whose account is `strongly corroborated by additional evidence supporting a guilty verdict ... generally is not found to be material,'" Rocha v. Thaler, 619 F.3d 387, 396 (5th Cir.2010) (quoting Sipe, 388 F.3d at 478), let alone on clear-error review and when the witness is an out-of-court declarant. Even if the net result of disclosing the McMahon notes to Brown would have been that the government would not have asked Glisan or Kopper to testify at all about what McMahon told them, that would have had essentially no impact on the government's case. Yet, it would have prevented Brown from making any use of the McMahon notes at trial, because they were otherwise inadmissible hearsay.23 Thus, although the McMahon notes are favorable evidence, disclosing them to Brown pre-trial would not have created a reasonable probability of a different outcome, even in conjunction with the suppressed Zrike testimony.

Turning to Zrike's testimony to the grand jury and SEC, Brown points to her statements that Merrill Lynch wanted to add a best-efforts clause but was "not successful in negotiating that [in] with Vinson & Elkins [Enron's outside counsel]." Zrike explained that Merrill Lynch was "trying to be creative to protect [itself], but they [the Enron legal team] kept coming back to the fact that it really had to be a true passage of risk...." She did not find it "nefarious [or] problematic" that Enron "would not put in writing an obligation to buy [the barges] back, to indemnify us[—]all those things were consistent with the business deal."

Those statements could have helped Brown by giving the defense an argument to counter the prosecution's position that the absence of a written "best efforts" agreement was evidence that there was no "best efforts" agreement at all. Brown could have pointed to Zrike's testimony to say that the reason the "best efforts" agreement was not in writing was that Enron's attorneys wanted a "true passage of risk." But that would have been of little marginal benefit to Brown, because Zrike already took the stand as a witness and gave testimony explaining that she believed the agreement was nothing more than a "best-efforts" agreement, and the prosecution successfully neutralized her testimony by arguing that she was unaware of Fastow's oral promise because Merrill Lynch's investment bankers kept her and the other lawyers out of the loop. Nothing in her allegedly suppressed testimony would have weakened the prosecution's successful argument on that point.

In sum, the favorable evidence that Brown points to is not, even cumulatively, sufficient to give us a "definite and firm conviction" that it establishes a substantial probability of a different outcome. There was considerable evidence of Brown's guilt. Trinkle testified that there was a promise during the conference call she listened in on; Glisan and Kopper testified about Fastow's statements to them that he promised to rebuy; Boyt testified that Boyle told him, immediately after the Fastow call, that Fastow promised a buyback during the call; Long testified that there was a promise as well; Merrill Lynch conducted no due diligence, consistent with a buyback promise; a number of contemporaneous emails and documents referred to a promise; there was in fact a buyback, at 15% return, exactly six months after Merrill Lynch bought the barges, just as some internal documents said would happen; Fuhs jokingly emailed Brown that he would re-buy the barges only if Brown gave him a buyback guarantee; and in an email Brown himself said Enron had made a promise to buy back.

Brown points to the divided panel in Brown I to argue that the evidence against him was relatively weak. It is true that the panel was divided on Brown's guilt, but that division was over whether a legally unenforceable oral promise could establish Brown's guilt, not whether there was an oral promise at all.24 The alleged Brady evidence in this appeal addresses only the latter issue—whether there truly was an oral promise to buy back or whether, instead, it was just a promise to use best efforts. It thus does not call the majority's holding in Brown I into question, and we have no authority to relitigate the issue that divided that panel. In short, the district court did not commit reversible error in holding that the Brady items, taken together, did not create a reasonable probability of a different outcome.

AFFIRMED.